Overview
Advances in genomics are creating new opportunities to understand the biology that require both systems modeling and bioinformatics. The ninth annual SysMod meeting will be a forum for discussion about the combined use of systems biology modeling and bioinformatics to understand biology, and disease. The meeting will take place in July 2025, during the 2025 ISMB/ECCB conference in Liverpool, UK. The meeting will feature two keynote talks and contributed presentations.
Topics
Methods
Dynamical modeling Flux balance analysis Logical modeling Network modeling Stochastic simulation …
Systems
Animals Bacteria Humans Plants Yeast …
Applications
Bioengineering Cancer Developmental biology Immunology Precision medicine …
Schedule
SysMod meeting:
11:20-13:00 | Session I: Computational Advances in Metabolic Modeling Moderator: Matteo Barberis, University of Surrey, United Kingdom |
---|---|
11:20 – 11:30 | Welcome and Introduction to SysMod 2025 Matteo Barberis University of Surrey, United Kingdom
The community of special interest (COSI) in systems modeling (SysMod) organizes annual one-day gatherings. In 2025 the meeting comprises three sessions that cover a broad variety of topics, beginning with metabolic modeling, followed by the afternoon session on multiscale modeling and concludes with inference of cellular processes. This year’s meeting features two keynote speakers, Ronan Fleming and Jasmin Fisher. The event is hosted by Chiara Damiani and Matteo Barberis on behalf of the eight COSI organizers. This brief talk introduces all speakers, organizers, and main topics of the 2025 meeting. |
11:30-12:10 | Keynote talk: Variational kinetics: a variational formulation of reaction kinetics Ronan Fleming School of Medicine, University of Galway, University Rd, Galway, Ireland |
12:10-12:30 | A dynamic multi-tissue metabolic reconstruction reveals interindividual variation in postprandial metabolic fluxes Lisa Corbeij, Natal van Riel, and Shauna O’Donovan
Genome-scale metabolic models (GEMs) are large network-based metabolic reconstructions that can predict the flux of numerous metabolites making them valuable for analysing metabolism across a wide variety of human tissues and microbial species. However, the steady-state assumption needed to solve these GEMs limits their utility to study disturbances in metabolic resilience. In this study, we embed GEMs of the liver, skeletal muscle, and adipocyte into the Mixed Meal Model (MMM), a physiology-based computational model describing the interplay between glucose, insulin, triglycerides and non-esterified fatty acids (NEFAs). We implement dynamically updating objective functions for each GEM, where cellular objective depends on the model-calculated insulin values. The MMM is simulated using a fixed-step ODE solver; at each time-step the exchange reaction bounds for glucose, triglycerides, and NEFAs in each GEM are updated according to the MMM outputs and flux balance analysis is used to determine the metabolic fluxes. The insulin-dependent objective function allowed the GEMs to accurately simulate the transition from glucose and NEFA secretion in the fasting state to nutrient storage post meal. Moreover, the dynamic tissue-specific GEMs also correctly simulated postprandial changes in metabolites like lactate and glycerol, that are not directly modulated by the differential equations. Personalised hybrid multi-tissue Meal Models, derived from meal response data reveal changes in tissue-specific flux associated with insulin resistance and liver fat accumulation. This research demonstrates the potential of merging GEMs with physiological models to deepen our understanding of metabolic dynamics, offering promising avenues for personalized medicine in metabolic disorders. |
12:30-12:50 | Decoding organ-specific breast cancer metastasis through single-cell metabolic modeling Garhima Arora and Samrat Chatterjee
Breast organotropism, the preferential metastasis of breast cancer cells to specific organs, remains a critical challenge but clinically significant phenomenon, with a limited understanding of the metabolic factors driving site-specific colonization. In this study, we employed genome-scale metabolic models (GSMMs) integrated with single-cell RNA sequencing data from patient-derived xenograft models to investigate the metabolic basis of breast cancer organotropism. We constructed 14 tissue-specific metabolic models from primary breast tumors and their corresponding metastatic sites in the liver, bone, and brain and systematically explored metabolic perturbations associated with disease progression. Our analysis revealed distinct metabolic adaptations in metastatic tissues, characterized by upregulation in lipid metabolism, vitamin and cofactor metabolism, and amino acid pathways, particularly in bone and brain metastases compared to the liver. Furthermore, flux-based comparisons of primary tumors predisposed to different metastatic destinations identified metabolic signatures predictive of organotropism. Using robust Metabolic Transformation Algorithm (rMTA), we simulated gene over-expression and knock-out strategies, identifying candidate metabolic genes capable of driving primary to metastatic phenotypes during breast organotropism. This systems-level approach not only advances our understanding of the metabolic determinants of breast cancer organotropism but also highlights potential metabolic targets for therapeutic intervention aimed at halting metastatic progression. |
12:50- 12:55 | Enzyme activation network facilitates regulatory crosstalk between metabolic pathways Sultana Al Zubaidi, Muhammad Ibtisam Nasar, Richard Notebaart, Markus Ralser, and Mohammad Tauqeer Alam
The metabolic network, the largest inter connected system in the cell, is constantly regulated by a range of regulatory interactions. To characterize metabolite-enzyme activatory interactions we reconstructed the cell-intrinsic enzyme-metabolite activation-interaction network (“activation network”) using the Saccharomyces cerevisiae metabolic model (Yeast9) as the basis. We integrated the Yeast9 metabolic network with the list of cross-species activating compounds from the BRaunschweig ENzyme DAtabase (BRENDA) database. The cell-intrinsic activation network comprises 1,499 activatory interactions involving 344 enzymes and 286 cellular metabolites. Although only 54% of yeast metabolic enzymes (344 out of 635) are intracellularly activated, these enzymes are distributed across nearly all pathways, underscoring the widespread role of activation in cellular metabolism. Notably, in 94% of pathways at least one initial reaction is intracellularly activated. These initial reactions are typically non-equilibrium, flux-generating steps that must be regulated to control the overall pathway flux. Moreover, our analysis shows that highly activating metabolites are predominantly essential, whereas highly activated enzymes tend to be non-essential for growth. Additionally, we find that activator metabolites are produced in fewer steps compared to non-activators, suggesting a streamlined synthesis for regulatory compounds. We further examined cross-pathway activation and found a significant degree of trans-activation, emphasizing the interconnected nature of cellular metabolism. This coordination ensures that metabolic pathways are selectively activated and dynamically adjusted to meet cellular demands. |
12:55- 13:00 | Cell-cycle dependent DNA repair and replication unifies patterns of chromosome instability Bingxin Lu, Samuel Winnall, Will Cross, and Chris Barnes
Chromosomal instability (CIN) is pervasive in human tumours and often leads to structural or numerical chromosomal aberrations. Somatic structural variants (SVs) are intimately related to copy number alterations but the two types of variant are often studied independently. Additionally, despite numerous studies on detecting various SV patterns, there are still no general quantitative models of SV generation. To address this issue, we develop a computational cell-cycle model for the generation of SVs from end-joining repair and replication after double-strand break formation. Our model provides quantitative information on the relationship between breakage fusion bridge cycle, chromothripsis, seismic amplification, and extra-chromosomal circular DNA. Given whole-genome sequencing data, the model also allows us to infer important parameters in SV generation with Bayesian inference. Our quantitative framework unifies disparate genomic patterns resulted from CIN, provides a null mutational model for SV, and reveals deeper insights into the impact of genome rearrangement on tumour evolution. |
13:00- 14.00 | Lunch Break |
14.00-16.00 | Session II: Systems biology and multiscale modeling Moderator: Chiara Damiani, Università degli Studi di Milano-Bicocca, Italy |
14:40-14:40 | Keynote talk: Virtual Tumours for Predictive Precision Oncology Jasmin Fisher University of Galway, Ireland
Cancer is a complex systemic disease driven by genetic and epigenetic aberrations that impact a multitude of signalling pathways operating in different cell types. The dynamic, evolving nature of the disease leads to tumour heterogeneity and an inevitable resistance to treatment, which poses considerable challenges for the design of therapeutic strategies to combat cancer. In this talk, I will discuss some of the progress made towards addressing these challenges, using the design of computational models of cancer signalling programs (i.e., virtual tumours). I will showcase a growing library of mechanistic, data-driven computational models, focused on the intra- and inter-cellular signalling in various types of cancer (namely triple-negative breast cancer, non-small cell lung cancer, melanoma and glioblastoma). These computational models are predictive and mechanistically interpretable, enabling us to understand and anticipate emergent resistance mechanisms and to design patient-specific treatment strategies to improve outcomes for patients with hard-to-treat cancers. |
14:40-15:00 | A community benchmark of off-lattice multiscale modelling tools reveals differences in methods and across-scales integrations Thaleia Ntiniakou, Othmane Hayoun-Mya, Marco Ruscone, Alejandro Madrid Valiente, Adam Smelko, Jose Luis Estragués Muñoz, Jose Carbonell-Caballero, Alfonso Valencia, and Arnau Montagud
The emergence of virtual human twins (VHT) in biomedical research has sparked interest in multiscale modelling frameworks, particularly in their application bridging cellular to tissue levels. Among these tools, off-lattice agent-based models (O-ABM) offer a promising approach due to their depiction of cells in 3D space, closely resembling biological reality. Despite the proliferation of O-ABM tools addressing various biomedical challenges, comprehensive and systematic comparisons among them have been lacking. This paper presents a community-driven benchmark initiative aimed at evaluating and comparing O-ABM for biomedical applications, akin to successful efforts in other scientific domains such as CASP. Enlisting developers from leading tools like BioDynaMo, Chaste, PhysiCell, and TiSim, we devised a benchmark scope, defined metrics, and established reference datasets to ensure a meaningful and equitable evaluation. Unit tests targeting different solvers within these tools were designed, ranging from diffusion and mechanics to cell cycle simulations and growth scenarios. Results from these tests demonstrate varying tool performances in handling diffusion, mechanics, and cell cycle equations, emphasising the need for standardised benchmarks and interoperability. Discussions among the community underscore the necessity for defining gold standards, fostering interoperability, and drawing lessons from analogous benchmarking experiences. The outcomes, disseminated through a public platform in collaboration with OpenEBench, aim to catalyse advancements in computational biology, offering a comprehensive resource for tool evaluation and guiding future developments in cell-level simulations. This initiative endeavours to strengthen and expand the computational biology simulation community through continued dissemination and performance-oriented benchmarking efforts to enable the use of VHT in biomedicine. |
15:00-15:20 | Multi-objective Reinforcement Learning for Optimizing JAK/STAT Pathway Interventions: A Quantitative System Pharmacology Study Nhung Duong, Tuan Do, Tien Nguyen, Hoa Vu, and Lap Nguyen Background and Aims: JAK/STAT cancer pathway-oriented treatment optimization poses challenges due to pathwaycomplexity, feedback loops, resistance development, andharmonization between efficacy, immunity preservation, andtoxicity minimization. We aimed to develop and utilize acomprehensive in silico framework based on multi-objectivereinforcement learning (MORL) to identify optimalintervention strategies targeting the JAK/STAT pathway.Methods: We constructed a multi-scale mechanistic modelintegrating JAK/STAT intracellular signaling dynamics(ODEs), tumor-immune cell interactions, adaptive resistanceevolution, and pharmacokinetics/pharmacodynamics/toxicityof inhibitors (JAKi, STATi, cytokine blockers). We trainedMORL agents (multi-objective PPO) employing this model asan environment to discover treatment schedules balancingfour objectives: tumor reduction, immune preservation,resistance prevention, and toxicity minimization.Pareto-optimal strategies were used.Results: MORL successfully identified a diverse set ofnon-dominated intervention strategies, revealing inherenttrade-offs. Distinct treatment paradigms emerged, such asan efficacy-focused strategy yielding ~28% tumor reductionbut incurring higher toxicity/resistance, contrasted with aresistance-prevention strategy achieving excellentresistance/toxicity scores (>0.94, >0.25) but limited tumorcontrol (~6.4%). Sensitivity analyses highlighted SOCS3regulation, STAT kinetics, and resistance parameters ascritical determinants of outcomes.Conclusion: This study introduces a MORL-empoweredframework for navigating complex therapeutic trade-offs inJAK/STAT targeting. Our findings reveal diverse optimalstrategies and underscore key biological factorsinfluencing treatment success, offering a computationalbasis for the rational design and personalization ofJAK/STAT-targeted therapies and showcasing the potential ofMORL in quantitative systems pharmacology. |
15:20-15:40 | Decoding CXCL9 regulatory mechanisms by integrating perturbation screenings with active learning of mechanistic logic-ODE models Bi-Rong Wang, Maaruthy Yelleswarapu, and Federica Eduati
Despite advances in immunotherapy, pancreatic cancer remains highly lethal due to an immunosuppressive tumor microenvironment characterized by immune exclusion. Inducing chemokines like CXCL9 with targeted therapy can promote cytotoxic T cell recruitment. However, systematic approaches are needed to better understand chemokine regulation and identify drug combinations upregulating CXCL9 expression. We combined logic-ODE models with wet lab experiments for two pancreatic cancer cell lines (AsPC1 and BxPC3), investigating CXCL9 responses to combinations of cytokines (IFNγ, TNFα) and inhibitors targeting JAK, IKK, RAS, MEK, or PI3K. Analyzing model parameters fitted to the screening data revealed both shared and cell line-specific mechanisms. To efficiently navigate the large space of possible drug combinations, we developed a pipeline for experimental design, integrating active learning with mechanistic modeling. In this pipeline, an acquisition function iteratively selects the most informative conditions from a pool of unseen drug combinations; these are added to the training data to update the model. We benchmarked different acquisition functions on in silico data: “greedy” selects conditions predicted to yield the highest CXCL9 levels, while “uncertainty” prioritizes those where the model is least confident. Both strategies outperformed random sampling: “greedy” most efficiently identified high-CXCL9 conditions, while “uncertainty” improved overall model generalizability. We are currently performing wet lab experiments to validate in silico predictions. Our framework demonstrates how active learning can be combined with dynamic logic-based models to accelerate the discovery of immunomodulatory drug combinations, offering a generalizable approach for hypothesis-driven experimental design in systems biology. |
15:40-16:00 | ARTEMIS integrates autoencoders and Schrödinger Bridges to predict continuous dynamics of gene expression, cell population and perturbation from time-series single-cell data Sayali Anil Alatkar
Cellular processes like development, differentiation, and disease progression are highly complex and dynamic (e.g., gene expression). These processes often undergo cell population changes driven by cell birth, proliferation, and death. Single-cell sequencing enables gene expression measurement at the cellular resolution, allowing us to decipher cellular and molecular dynamics underlying these processes. However, the high costs and destructive nature of sequencing restrict observations to snapshots of unaligned cells at discrete timepoints, limiting our understanding of these processes and complicating the reconstruction of cellular trajectories. To address this challenge, we propose ARTEMIS, a generative model integrating a variational autoencoder (VAE) with unbalanced Diffusion Schrödinger Bridge (uDSB) to model cellular processes by reconstructing cellular trajectories, reveal gene expression dynamics, and recover cell population changes. The VAE maps input time-series single-cell data to a continuous latent space, where trajectories are reconstructed by solving the Schrödinger bridge problem using forward-backward stochastic differential equations (SDEs). A drift function in the SDEs captures deterministic gene expression trends. An additional neural network estimates time-varying kill rates for single cells along trajectories, enabling recovery of cell population changes. Using three scRNA-seq datasets—pancreatic β-cell differentiation, zebrafish embryogenesis, and epithelial-mesenchymal transition (EMT) in cancer cells—we demonstrate that ARTEMIS: (i) outperforms state-of-art methods to predict held-out timepoints, (ii) recovers relative cell population changes over time, and (iii) identifies “drift” genes driving deterministic expression trends in cell trajectories. Furthermore, in silico perturbations show that these genes influence processes like EMT. The code for ARTEMIS: https://github.com/daifengwanglab/ARTEMIS. |
16:00- 16.40 | Coffee Break |
16.40-18.00 | Session III: Analysis of single cells and and inference of cellular processes Moderator: |
16:40-17:00 | Calibrating agent‐based models of colicin-mediated inhibition in microfluidic traps using single-cell time-lapse microscopy Ati Ahmadi, Samantha Schwartz, and Brian Ingalls Bacterial antagonism plays a key role in shaping microbial communities in heterogeneous environments such as the gut or soil. Investigation of the spatiotemporal dynamics of these interactions helps reveal how spatial structure influences competition, survival, and resistance. In this study, building on our previous work, we present an integrated pipeline that combines high- resolution time-lapse microscopy, advanced image processing, and agent-based modeling to quantitatively characterize the dynamics of colicin-mediated killing. We focus on a system involving two E. coli strains: DH5α, engineered to produce the toxin colicin Ib, and K12, which expresses the cirA receptor, making it susceptible to the toxin. We cultured the two strains in more than a dozen microfluidic traps measuring 80 × 80 μm under a constant media flow. Phase-contrast and fluorescence monolayer images were acquired every 3 minutes over an 18-hour period. As the chambers filled, cells grew and divided until reaching a steady state, in which new daughter cells displaced older cells that exit through the open mouth of the trap. Using in-house image processing techniques, including segmentation and single-cell tracking—we extracted cell trajectories, growth rates, and other relevant features at single-cell resolution. From this data, we calculated a proximity-based exposure score for each susceptible cell by averaging its distance to the all attacker over time, weighted exponentially to reflect the localized effect of colicin. Classification by logistic regression revealed that this measure is a strong predictor of cell death, confirming the importance of spatial proximity in determining toxin exposure. We developed an agent-based model to simulate colicin Ib on susceptible cells. In the model, colicin is released by attacker cells and spreads through the environment via diffusion and undergoes degradation. The local concentration of colicin, C(x,t), is governed by a partial differential equation accounting for production, degradation, diffusion. A susceptible cell is programmed to die when the local colicin concentration C(x, t) exceeds a threshold. This formulation captures spatial gradients in toxin levels due to the interplay between secretion and rapid degradation. To quantitatively link our experimental observations with the agent-based model, we use an average exponential over lifetime feature as a summary statistic for model calibration. Specifically, for each susceptible cell, we compute an effective exposure score. We calibrated the production rate P and degradation rate δ using Sequential Monte Carlo based Approximate Bayesian Computation. This involved comparing the simulated exposure scores and other spatial features to experimental data, aiming to minimize the difference between them. The result is a set of parameters that best match the spatial killing patterns observed in the experiments, providing a foundation for characterizing system behavior under different spatial configurations and making predictions about outcomes in untested conditions. By combining high-resolution single-cell imaging with a well-calibrated computational model, we’ve quantitatively described how colicin-mediated killing depends on spatial organization. This integrated approach offers insights into how toxins spread and act in structured environments and opens new possibilities for engineering microbial communities, particularly in designing antimicrobial strategies. For example, by understanding how colicin concentration gradients emerge from spatial arrangements, one can design microbial consortia where toxin-producing strains are strategically positioned to suppress susceptible competitors without removing beneficial neighbors. This spatial structuring could be applied in the design of synthetic biofilms or encapsulated probiotics, where controlled positioning of antagonistic strains enhances stability and function. This aligns with recent efforts to leverage antagonism for shaping microbial ecosystems and provide a quantitative framework for engineering spatially organized consortia with desired competitive outcomes. |
17:00-17:20 | Inferring metabolic activities from single-cell and spatial transcriptomic atlases Erick Armingol, James Ashcroft, Magda Mareckova, Martin Prete, Valentina Lorenzi, Cecilia Icoresi Mazzeo, Jimmy Tsz Hang Lee, Marie Moullet, Christian Becker, Krina Zondervan, Omer Ali Bayraktar, Luz Garcia-Alonso, Nathan E. Lewis, and Roser Vento-Tormo
Metabolism is fundamental to cellular function, supporting macromolecule synthesis, signaling, growth, and cell-cell communication. While single-cell and spatial metabolomics technologies have advanced, large-scale applications remain challenging. In contrast, transcriptomics provides vast datasets to infer metabolic states. Here, we present scCellFie, a computational tool that predicts metabolic activities from transcriptomic data at single-cell and spatial resolutions. scCellFie enables scalable analysis of large cell atlases, leverages metabolic tasks for interpretable results, and includes modules for identifying metabolic markers, condition-specific changes, and cell-cell communication. We applied scCellFie to ~30 million human cells, generating a comprehensive metabolic atlas across organs while demonstrating our tool’s scalability. Additionally, we used scCellFie to study the human endometrium, the uterine lining that undergoes substantial remodeling throughout the menstrual cycle due to sex hormones, and identified cell type-specific metabolic programs supporting cyclical changes. Epithelial cells exhibited metabolic regulation covering pathways supporting proliferation and mitigating oxidative stress. Endometrial diseases, including endometriosis and endometrial carcinoma, often arise from metabolic dysregulation. By inspecting eutopic endometrium from donors with endometriosis, we identified altered metabolic programs that likely drive atypical proliferation and inflammation of the distinct cell types. In endometrial carcinoma, malignant cells displayed metabolic rewiring, including increased glucose-to-lactate conversion and dysregulated kynurenine and estrogen signaling. These shifts suggest shared mechanisms promoting aberrant proliferation and may reveal therapeutic targets. Together, our findings demonstrate scCellFie as a scalable, interpretable tool for characterizing metabolism in health and disease. By linking metabolic functions to cellular processes, scCellFie provides deeper insights into metabolic regulation across diverse biological systems. |
17:20-17:40 | Spatiotemporal Variational Autoencoders for Continuous Single-Cell Tissue Dynamics Koichiro Majima and Teppei Shimamura
Single-cell spatial genomics provides unprecedented molecular insights, yet it still struggles to track both the spatial and temporal progression of tissues under native conditions. Experimental constraints and destructive sampling yield discrete snapshots rather than the continuous record required to fully understand how cells organize and function over time. Optimal transport (OT) approaches have attempted to bridge these snapshots across different assays, but they typically rely on unimodal data, scale poorly, and oversimplify the complexity of ongoing morphogenetic events. We introduce a spatiotemporal variational autoencoder (VAE) that models the continuous evolution of tissue pixels—capturing dynamic changes in both spatial location and gene-expression patterns. By embedding these pixels into a latent space governed by a learned dynamics network, our method reveals how tissues grow, reorganize, and express key genes over time. At each time point, behavioral parameters are decoded from the latent state via a neural decoder, quantifying probabilities of growth, disappearance, seeking, displacement, attraction, and clustering. A differentiable growth module further refines this process by modeling region appearance and disappearance, allowing gradient-based optimization of tissue occupancy patterns. We demonstrate the power of our approach by analyzing a mouse embryogenesis dataset. The model uncovers unobserved developmental trajectories, pinpoints morphogenetic transitions, and aligns coronal sections at scale—capabilities that standard OT-based methods find intractable. These results highlight how a spatiotemporal VAE can reconstruct the story of tissue formation in both forward and backward directions, opening up new avenues to interpret single-cell and spatial data as a cohesive, dynamic narrative rather than disjointed snapshots. |
17:40-17:45 | Computational Modeling of Shortening and Reconstruction of Telomeres Marek Kimmel, Marie Doumic, Leonard Mauvernay, and Teresa Teixeira
We discuss a stochastic model of growth of a cell population of cultured yeast cells with gradually decaying chromosome endings called the telomeres, as well as models of telomere reconstruction using the so-called ALT (alternate lengthening of telomeres) Mechanism. Telomeres play a major role in aging and carcinogenesis in humans. Our models correspond in part to the experiments of one of us (Teixeira). For telomere shortening, we modify the method of Olofsson and Kimmel, who considered properties of the branching process of telomere shortening; this leads to consideration of a random walk on a two-dimensional grid. We derive an integral equation for the probability generating functions (pgf’s) characterizing the dynamics of shortening of telomeres. We find that the general solutions have the form of exponential polynomials. Stochastic simulations lead to interesting and non-obvious effects if cell death is included. We further consider more complex models, involving cell death and the ALT mechanism. of telomere reconstruction. These are based on our works and are intended to address the experiments in Kockler et al. (2021). In one version of the ALT Mechanism model, we consider expectations conditional on non-extinction, since only a fraction of ALT telomeres is stably elongated (see Kockler et al. 2021). As a conclusion, the multitype branching processes produce realistic prediction concurrent with complex biological experiments involving telomeres. Our aim is to use the models for longer-term prognoses for human telomeres. |
17:45-17:50 | TFvelo: gene regulation inspired RNA velocity estimation Jiachen Li, Xiaoyong Pan, Ye Yuan, and Hong-Bin Shen
RNA velocity is closely related with cell fate and is an important indicator for the prediction of cell states with elegant physical explanation derived from single-cell RNA-seq data. Most existing RNA velocity models aim to extract dynamics from the phase delay between unspliced and spliced mRNA for each individual gene. However, unspliced/spliced mRNA abundance may not provide sufficient signal for dynamic modeling, leading to poor fit in phase portraits. Motivated by the idea that RNA velocity could be driven by the transcriptional regulation, we propose TFvelo, which expands RNA velocity concept to various single-cell datasets without relying on splicing information, by introducing gene regulatory information. Our experiments on synthetic data and multiple scRNA-Seq datasets show that TFvelo can accurately fit genes dynamics on phase portraits, and effectively infer cell pseudo-time and trajectory from RNA abundance data. TFvelo opens a novel, robust and accurate avenue for modeling RNA velocity for single cell data. |
17:50-18:00 | Closing Remarks Chiara Damiani Università degli Studi di Milano-Bicocca, Italy
This concluding talk aims to briefly discuss the diversity of topics presented at the “Computational Modeling of Biological Systems” (SysMod) COSI track. This diversity illustrates the importance of the field and the broad range of applications in systems biology and disease. Then, forthcoming meetings of interest will be announced, and the three poster awards will be delivered as a closing event. |
Poster presentation:
10:00-11:20 and 16:00-16:40 | Session A: July 21, 2025 |
---|---|
A-389: Time Domain Simulation of Signaling in Biochemical Chain Reactions of Metabolic Pathways
Farzaneh Taringou, ACE – Advanced Computational Engineering GmbH, Germany
Thomas Weiland, ACE – Advanced Computational Engineering GmbH, Technische Universitaet Darmstadt, Germany
This work presents the preliminary results of modeling, simulation and visualization of time domain interactions of the signaling on the metabolic pathways upon excitation with single, simultaneous or sequential triggers. The model is predictive of the flow of the forward and backward biochemical reactions in time domain. The modeling approach is developed to capture the dynamic behavior and is not intended to reflect the mechanical or material properties of cells or organoids involved. We concentrated on the pathways of Tricarboxylic acid cycle (TCA), electron transport chain (ETC) and Heme biosynthesis. TCA and ETC are entirely routed inside mitochondrial region, whereas the pathway of heme biosynthesis runs halfway in cytosol and halfway in the mitochondria. We model the biochemical pathways as electrical signal carriers. The electrochemical signaling is projected onto electromagnetic wave propagation. The wave equation is solved numerically in time domain employing Finite Integration Technique. The model is excitable through user-defined electrical signals at specified input ports. Scenarios with several metabolic excitations simultaneously or sequentially can be solved and illuminated using high-end visualization techniques and CAD models. Partial deactivation of delta-aminolaevulinic acid dehydratase of the pathway of heme-biosynthesis has been modeled and simulated. The results are compared with those of a functional pathway. The interaction of the backward signals with the interconnected pathways of ETC and TCA has been visualized. The backward signaling is a crucial consideration which could predict the rate and sites of congestions and substrate accumulations which manifest themselves as chronic or acute phases of a disease. |
|
A-391: Spectroscopic analysis, kinetic mechanism, computational docking, and molecular dynamics of active metabolites from the fruits of Hovenia dulcis Thunb. as potential inhibitors of PTP1B and α-glucosidase
Jeong Ah Kim, Vessel-Organ Interaction Research Center, VOICE (MRC), College of Pharmacy, Kyungpook National University, Daegu, 41566, South Korea
Byung Sun Min, College of Pharmacy, Drug Research and Development Center, Daegu Catholic University, Gyeongbuk 38430, South Korea
Sooyeun Lee, College of Pharmacy, Keimyung University, Daegu 42601, South Korea
Hye Kyoung Sung, Department of Biology, York University, 4700 Keele Street, Toronto, ON M3J1P3, Canada
Hovenia dulcis Thunb. is a flowering plant of the Rhamnaceae family that is native to East Asia, particularly China and Korea. From the fruits of H. dulcis Thunb., two new 2,3-dihydrobenzofuran derivatives, named hovedulcates A and B (14 and 16), as well as 20 known compounds (1–13, 15, and 17–22) were isolated. The structures of the isolated compounds were elucidated through a combination of NMR, HR-MS, and ECD spectroscopy analyses, and reported data in the literature. The inhibitory activities of the isolated compounds (1–22) against protein tyrosine phosphatase 1B (PTP1B) and α-glucosidase were assessed. Notably, new compounds 14 and 16 exhibited potent inhibitions of both PTP1B and α-glucosidase. The enzyme kinetics analysis was conducted to understand the inhibition mode and inhibitory constants of compounds 14 and 16 for the inhibition of PTP1B and α-glucosidase. Additionally, molecular docking simulations revealed the binding mechanisms and interactions of these compo |
|
A-393: Repurposing whole-cell modelling tools
Yin Hoon Chew, University of Birmingham, United Kingdom
Whole-cell models (WCMs) are genome-scale, data- and knowledge-driven computational models of cells. They integrate multi-omics, biochemical and physiological data into mechanistic representations of intracellular events; and enable dynamic simulation of cellular behaviours such as growth. I previously collaborated with others to develop a suite of Python APIs to semi-automate WCM development. These APIs include a schema for organising data into structured knowledge base and a language (that comes with a compatible simulator) for representing WCMs. Even though these APIs were part of a pipeline for whole-cell modelling, some can be repurposed for more general use. I will present these tools and their features, for those interested in whole-cell modelling, as well as researchers working with data and models about cellular processes. |
|
A-395: A novel framework leveraging multi-omics integration, knowledge graph and topological data analysis to analyze gene regulatory networks to decipher plant responses to multiple stresses
Silvia Bottini, INRAE, Université Côte d’Azur, CNRS, Institut Sophia Agrobiotech, Sophia-Antipolis, France, France Maxime Multari, INRAE, Université Côte d’Azur, CNRS, Institut Sophia Agrobiotech, Sophia-Antipolis, France, France
Mathieu Carriere, Data Shape Team, Centre Inria d’Université Côte d’Azur, France, France
Xavier Amorós-Gabarrón, INRAE, Université Côte d’Azur, CNRS, Institut Sophia Agrobiotech, Sophia-Antipolis, France, Franc
Aurelien Dugourd, EMBL-EBI, Wellcome Genome Campus, Cambridgeshire, United Kingdom, United Kingdom
Global population is rapidly increasing, representing a major challenge for food supply, exacerbated by climate change and environmental degradation. Most of those depend on agriculture, however, plant health is threatened by various stressors. Although how plants respond to each of these individual stresses is well studied, little is known about how they respond to a combination of many of these bio-aggressors occurring together. To tackle this question, we analyzed the transcriptome response of tomato to six distinct pathogens with our novel integrative approach, HIVE. Then we used TomTom, a knowledge graph which we developed by gathering molecular interactions from nine databases, to extract a comprehensive gene regulatory network (GRN) to study how multi-pathogens stress response is coordinated. By estimating the transcription factor activity, we identified 43 TFs responding either specifically to one bio-aggressor or to multiple ones. The analysis of the GRN with a topological data analysis approach allowed to identify 18 clusters of transcription factors hierarchically organized in four main configurations depending on TF activity and shared targets. Finally, we found four ERF hubs that cooperatively orchestrated the tomato response to multiple pests. Our tools allowed us to study the complex molecular reprogramming in tomatoes upon interaction with a wide range of biotic agents, providing methods scalable to other phytopathosystems. |
|
A-397: Evaluating Cell Segmentation Methods for Spatial Transcriptomics: Challenges and Comparative Insights
Daria Lazic, Genome Biology Unit, EMBL, Heidelberg, Germany, Germany
Wolfgang Huber, Genome Biology Unit, EMBL, Heidelberg, Germany, Germany
Accurate assignment of transcripts to cells in imaging-based spatial transcriptomics platforms, such as 10x Xenium, remains a major challenge due to cellular overlap, the absence of nuclei in tissue sections, lack of pan-cellular membrane stains, heterogeneous transcriptional activity, transcript diffusion, and irregular cell morphologies. Segmentation methods that rely solely on nuclear or membrane stains are prone to missing cells and misassigning transcripts, leading to downstream errors such as incorrect cell type classification, spurious expression correlations, false-positive cell-cell interactions, and misinterpretation of transient cell states or cell transitions. In this study, we first performed a broad literature screen of existing segmentation methods, comparing them across criteria including input requirements (e.g. prior information such as scRNA-seq data or nuclear segmentation), memory usage, and runtime. We then carried out a more detailed evaluation of three selected methods – BIDCell, ProSeg and Segger – and compared their results against manufacturer-provided segmentation masks from 10x Genomics. Our analysis focused on baseline cell-level metrics such as morphological cell features and transcript/gene counts per cell, as well as biological plausibility assessed via expression purity, neighborhood contamination upon label transfer, and correlation to publicly available scRNA-seq datasets. Our initial insights underscore the need for benchmark metrics tailored to specific biological questions. In particular, we argue that accurate cell shape reconstruction and correct transcript assignment should be evaluated as distinct objectives due to transcript diffusion. Importantly, we observe that methods treating cells as 2D objects inherently introduce neighborhood contamination, particularly in densely packed or overlapping cellular regions. |
|
A-399: RNA Velocity Comparison Method Based on Contrastive Learning for Cross-Condition Analysis
Recent technological advances have enabled the acquisition of single-cell RNA sequencing (scRNA-seq) data, facilitating gene expression analysis at single-cell resolution. However, scRNA-seq data provide only static snapshots of gene expression at a single time point. Such data does not provide direct temporal information about dynamic processes including cell differentiation. RNA velocity analysis provides temporal information by estimating rates of change in gene expression based on the ratios of unspliced pre-mRNA to spliced mRNA, allowing the inference of cell differentiation trajectories from data captured at single time points. Existing RNA velocity methods are limited to within-sample analyses presenting limitations when comparing cell differentiation trajectories under two conditions (e.g. experimental vs. control). Since each sample is embedded in an independent low-dimensional space, there is no shared embedding space, making it impossible to directly compare the direction and magnitude of RNA velocity differences across conditions. We propose a novel framework that leverages contrastive learning to construct a shared embedding space for datasets from different conditions. Our method enables visualization and quantification of differences in cell differentiation trajectories based on gene expression dynamics within a unified embedding space. This approach allows the direct comparisons of RNA velocity direction and magnitude across experimental conditions, providing a framework for quantitatively evaluating how genetic perturbations alter differentiation trajectories. |
|
A-401: Walking the BooLEVARD: Quantifying Signal Transduction in Boolean Models
Boolean models are a powerful resource for analyzing the dynamic behavior of biological systems. However, their binary nature poses limitations when it comes to capturing graded phenomena like signal intensity or protein activity levels. While some current tools allow for exploration of signal transduction pathways, they often rely on simplifications that reduce the analytical depth and accuracy due to computational constraints. Here we present BooLEVARD, a Python package developed to quantitatively evaluate the number of activation and repression pathways in Boolean models. By focusing on non-redundant causal paths, BooLEVARD enables a more nuanced interpretation of how signals traverse across complex networks.This method allows for a more precise representation of signal intensity within a discrete logic framework. As a case study, we apply BooLEVARD to a Boolean model of cancer metastasis to investigate how specific pathways contribute to cell-fate decisions. Our results highlight the tool’s capacity to pinpoint key signaling events and improve our understanding of cellular responses in health and disease. BooLEVARD is open-source and avaliable at https://github.com/farinasm/boolevard. |
|
A-403: A PCA-based synthetic tool for local sensitivity analysis of chemical reaction networks
Giorgia Biddau, Università degli Studi di Genova, Italy Giacomo Caviglia, Università degli Studi di Genova, Italy
Michele Piana, Università degli Studi di Genova ; LISCOMP Lab, IRCCS Ospedale Policlinico San Martino, Italy Sara Sommariva, Università degli Studi di Genova, Italy
Many phenomena related to cell signalling and underlying diseases such as cancer have been successfully modelled using chemical reaction networks (CRNs). Since realistic CRNs involve numerous proteins and reactions, precise calibration of kinetic parameters is essential, but would require an unrealistic amount of data. Here, we propose a mathematical tool based on the calculation of so-called statistical sensitivity indices (SSIs) to rank the parameters according to their sensitivity [3], identify key reactions in CRNs and highlight the impact of mutations on cell dynamics.
From a mathematical viewpoint, the computation of SSIs involves two steps: the analytically computation of the local sensitivity matrix, which comprises the partial derivative of the proteins’ concentration at equilibrium with respect to the kinetic parameters, and the application of Principal Component Analysis to a rescaled version of the sensitivity matrix. Our approach was validated on a CRN modelling signal transduction of colorectal cells [2, 4], comprising 419 proteins involved in 10 interacting pathways, resulting in a total of 850 reactions. The SSIs enabled the quantification of the impact of four prevalent colorectal cancer mutations and identified triggered subpathways. Furthermore, we found that in a combined therapy comprising two target drugs, Dabrafenib (DBF) and Trametinib (TMT), a finer calibration of DBF dosage should be preferred. [1] Krishnan, J., et al., J. Math. Biol., 2020[2] Sommariva S. et al., Sci. Rep., 2021[3] Biddau G. et al., Sci. Rep., 2024[4] Yu P. et al., Isr. J. Chem., 2018[5] Tortolina L. et al., Oncotarget, 2015 |
|
A-405: CHARMM parameterization of the isomers and analogs of retinal from jumping spider Rhodopsin-1 using the Force Field Toolkit
Gabriel J. Olguín-Orellana, Universidad Católica del Maule, Laboratorio de Bioinformática y Química Computacional, Chile, Chile Victor E. Rojas-Pérez, Universidad Católica del Maule, Laboratorio de Bioinformática y Química Computacional, Chile, Chile Cristobal Avendaño, Universidad Católica del Maule, Laboratorio de Bioinformática y Química Computacional, Chile, Chile Daniel Bustos, Universidad Católica del Maule, Laboratorio de Bioinformática y Química Computacional, Chile, Chile Mauricio Bedoya, Universidad Católica del Maule, Centro de Investigación de Estudios Avanzados del Maule, Talca, Chile, Chile Reynier Suardíaz, Universidad Complutense de Madrid, Departamento de Química Física, Facultad de Ciencias Químicas, Madrid, España, Spain Erix W. Hernández-Rodríguez, Universidad Católica del Maule, Laboratorio de Bioinformática y Química Computacional, Chile, Chile
Rhodopsins are photosensitive proteins that play a fundamental role in phototransduction and are invaluable in optogenetics, allowing for neuronal modulation through light. Jumping spider Rhodopsin-1 (JSR1) stands out for its bistability, which makes it especially promising for this field of research. Upon absorbing photons, rhodopsin’s chromophore, retinal, undergoes isomerization, triggering conformational changes in JSR1 that drive signal transduction. Retinal can exist in several isomeric forms—9-cis, 11-cis, 13-cis, and all-trans—as well as in 3,4-didehydro-retinal analogs, including 11-cis and all-trans, each with distinct photochemical properties essential for specificity in light interaction.
In this study, we developed CHARMM-compatible parameters for these six forms of retinal using the Force Field Toolkit (ffTK) in VMD. The parameterization comprised six steps: molecular model construction, assignment of van der Waals interactions, geometric optimization, charge derivation, bond and angle refinement, and dihedral calibration. We performed Hartree-Fock and MP2 calculations with the 6-31G* basis set for geometric optimization, followed by the derivation of partial charges to accurately reproduce solute-water interactions. Bond and angle parameters were refined through Hessian analysis at the quantum level, capturing their potential energy surfaces. Dihedral terms were also optimized by scanning the potential energy surface and carrying out iterative refinements.* The resulting parameters enable molecular dynamics simulations of JSR1 with each form of retinal, offering insights into the structural dynamics underpinning JSR1’s function. Integrating these parameter sets into CGenFF will allow for accurate computational modeling of the retinal isomers and 3,4-didehydro-retinal analogs, advancing our understanding of their roles in light-mediated neuronal control. |
|
A-407: Influence of Depression Elevation on the Force Distribution of Tibia Plateau
Mahziyar Darvishi, University of Toronto, Toronto, Canada, Canada
Timothy Burkhart, University of Toronto, Toronto, Canada, Canada
Tibial plateau fractures (TPFs) are complex joint injuries, comprising 1–2% of all fractures, with an annual incidence of 10.3 per 100,000. Schatzker II split-depression fractures are the most common subtype, representing 35% of TPFs. These injuries involve a lateral plateau split with central depression, typically managed with bone grafting and plate fixation. Improper surgical correction—either under- or over-correction—can significantly alter joint contact mechanics, yet the biomechanical effects of such variations remain understudied. This study aims to quantify the biomechanical impact of under- and over-correction in Schatzker II fractures using the finite element method (FEM). CT-based 3D models of the tibia and femur were reconstructed in 3D Slicer and refined in 3-matic to define anatomical and graft regions. Three models were developed: normally corrected, 2.5 mm over-corrected, and 2.5 mm under-corrected. Finite element meshes were generated in ABAQUS CAE, with all materials assumed to be homogeneous, isotropic, and linearly elastic. A 2 mm displacement was applied to the proximal femur while fixing the distal tibia. Over-correction increased lateral stress and displacement in the tibial cartilage and subchondral bone, indicating a higher risk of joint overload. In contrast, under-correction led to stress and displacement shifts toward the medial and antero-medial regions, with the greatest subchondral displacement observed, raising concerns about instability and medial strain. Additionally, over-correction raised distal tibial reaction forces, suggesting altered load transmission and potential gait abnormalities. These findings underscore the importance of precise intraoperative correction to maintain natural load distribution and minimize long-term joint degeneration. |
|
A-409: Comparative Systems Biology Reveals Divergent and Shared Host Responses to COVID-19 and Vaccination
Rima Hajjo, Compoutational Chemical Biology, School of Pharmacy, Al-Zaytoonah University of Jordan, Amman, Jordan, Jordan
Coronavirus Disease 2019 (COVID-19) and its vaccines elicit complex, multi-system host responses that can be elucidated through integrative computational biology. A comprehensive multi-omics informatics platform was applied, combining transcriptomics, network biology, pathway enrichment, and chemogenomics, to compare the biological effects induced by COVID-19 infection and mRNA vaccination. Transcriptomic analyses of COVID-19 implicated MAPK, JAK-STAT, fibronectin, and coagulation pathways, leading to the prioritization of von Willebrand factor (VWF) as a therapeutic target. These findings supported host-targeted drug repurposing and synergistic treatment strategies, such as combining glucocorticoids with long-acting beta-2 agonists (LABAs), to mitigate severe disease outcomes. In contrast, analysis of mRNA vaccines, particularly BNT162b2, revealed activation of interferon and cytokine signaling along with STAT1, STAT2, and IRF-driven transcriptional programs. While these responses are central to protective immunity, they were also associated with potential adverse effects involving calcium homeostasis and hormone-related signaling. Adverse events such as myocarditis, pericarditis, and menstrual irregularities were mechanistically linked to transcription factor activation and pro-inflammatory mediators. Connectivity Map analysis reinforced these associations by identifying cardiac and hormonal perturbagens that mirrored the vaccine-induced transcriptional profile. Comparative evaluation with live or live attenuated vaccines, such as smallpox and anthrax, revealed convergent host vulnerabilities, particularly in young males, suggesting shared immunological response pathways across vaccine platforms. This integrative systems biology approach provides mechanistic insight into both protective and adverse host responses and offers a scalable framework for optimizing vaccine safety, therapeutic targeting, and public health strategy across emerging infectious diseases. |
|
A-411: Agent-Based Modeling of Macrophage-Fibroblast Interactions in Response to Pathogens and Implanted Biomaterials
The interaction between macrophages and fibroblasts plays a key role in regulating inflammation, tissue repair and fibrosis in response to infection, injury or other pathological conditions. Macrophages, as central immune cells, exhibit a range of functional phenotypes depending on the signals that they receive and their environment. This polarization shifts macrophages along a continuum range between pro-inflammatory (M1) and pro-healing (M2) states, and is regulated by cytokines that activate and recruit other cells, including fibroblasts, thereby shaping the tissue response.
Fibroblasts are essential for tissue remodeling. They produce extracellular matrix (ECM) components such as fibrin and collagen, and can encapsulate implanted biomaterials within a fibrotic layer, isolating them from surrounding tissue. Macrophages and fibroblasts interact in both healthy and diseased tissues, primarily through signaling involving colony-stimulating factor 1 (CSF1) and its receptor CSF1R. In this context, we propose an agent-based (AB) modeling approach to simulate the response of relevant cell populations to external stimuli, such as pathogens or implanted biomaterials. The model explicitly describes macrophage phenotype transitions and their interactions with fibroblasts. To support this, we first evaluate existing ordinary differential equation (ODE) and AB models that simulate macrophage dynamics, validating them against in vitro experimental data. Results show that the AB model aligns more closely with experimental observations than the ODE models. Based on these findings, we develop an extended AB platform incorporating fibroblasts and fibrosis following biomaterial implantation, enabling realistic simulation of in vivo immune dynamics and showing strong agreement with validation data. |
|
A-413: Coxmos: Addressing the Challenges of Survival Analysis for Multi-Omic Data
Pedro Salguero, Department of Applied Statistics, Operations Research and Quality, Universitat Politècnica de València, Spain
Anabel Buendía, Department of Applied Statistics, Operations Research and Quality, Universitat Politècnica de València, Spain
Sonia Tarazona, Department of Applied Statistics, Operations Research and Quality, Universitat Politècnica de València, Spain
Survival analysis is a fundamental tool in many fields, including medicine, with the Cox proportional hazards model being the most popular survival model. However, the analysis of high-dimensional (HD) datasets, such as omic or multi-omic (MO) data, presents significant challenges due to the dimensionality and multicollinearity. Machine learning techniques can handle these limitations, but they often lack interpretability. To overcome these issues, we adapted Partial Least Squares Regression (PLS) methods for survival (PLS-Cox), since PLS models benefit from multicollinearity, manage HD data and are more interpretable.
Our R package Coxmos incorporates both classical Cox models with stepwise or Elastic Net variable selection, and PLS-based survival models for single-omic datasets (sPLS-ICOX, sPLS-DRCOX and sPLS-DACOX), as well as for MO datasets, with two different approaches: single-block (concatenating omics) and multi-block. In addition, Coxmos provides tools for model evaluation and comparison through metrics such as AIC, C-index, Brier Score and AUC, along with cross-validation for hyperparameter optimization, visualization options and interpretation in terms of the original variables. We applied Coxmos to real HD and MO datasets to evaluate the different survival modelling strategies. The results showed that sPLS-COX models generally outperformed classical approaches according to metrics and computational efficiency, with multi-block methods achieving better results in specific contexts. In conclusion, Coxmos provides a robust and interpretable framework for survival modelling, while highlighting the importance of strategy selection based on dataset characteristics and analysis goals. |
|
A-415: Personalizable Hybrid Modelling of Postprandial C-peptide
Max de Rooij, Eindhoven University of Technology, Netherlands
Natal van Riel, Eindhoven University of Technology, Netherlands
Shauna O’Donovan, Eindhoven University of Technology, Netherlands
Universal differential equations (UDEs) are an emerging approach combining physiology based mathematical models (PBMMs) with artificial neural networks (ANNs). However, in many biological problems, there is considerable inter-sample variation. This variation may arise from measurement noise, or due to differences in genetic background or acquired exposures that are indicative of pathological conditions in the development of a disease state of interest. For this reason, it is important that modelling approaches can adequately capture and quantify this data heterogeneity, thereby improving our understanding of the underlying disease. Conventional UDEs cannot accommodate data heterogeneity, limiting direct application to biological modelling problems. We propose a conditional UDE (cUDE) framework which introduces a tunable conditioning parameter to the ANN in the model to account for individual differences. In this way, the biases and weights of the ANN learn the population level response while the conditioning parameter captures inter-individual variation.
We applied cUDEs to learn a model of c-peptide production, a marker of insulin secretion, in individuals with varying glucose tolerance status. The trained cUDE model accurately describes postprandial c-peptide levels across the whole population, outperforming a conventional UDE. Furthermore, the tunable individual parameter produced a strong correlation with independently measured hyperglycemic clamp indices, the gold standard measure of beta-cell capacity (rho = -0.81). We demonstrated that this mathematical model can reliably quantify insulin production capacity from oral glucose tolerance tests data, providing a valuable surrogate index for use in precision healthcare in diabetes and metabolic diseases. |
|
A-417: NephrESA: Machine learning-informed mechanistic mathematical modelling to optimize anaemia management in chronic kidney disease and cancer
Anemia is a condition characterized by a reduced number of red blood cells (erythrocytes), which contain the oxygen-carrying protein hemoglobin (Hb). It is a common and multifactorial complication in chronic diseases such as chronic kidney disease and cancer. In CKD and cancer, anemia commonly arises from reduced erythropoietin production, inflammation, or treatment effects, and is often treated with erythropoiesis-stimulating agents (ESAs). However, in 49 dialysis patients, we observed hemoglobin fluctuations despite guideline-based ESA use, likely due to patient-specific differences. To reduce hemodynamic stress from ESA treatment, we developed a machine learning–informed, mechanistic multiscale modeling framework to support clinical decision-making in anemia management, using CKD as a use case. We calibrated the model using longitudinal data from over 200 healthy individuals and 1,500 CKD patients, including data on hemoglobin levels and ESA administration. Parallel tree boosting algorithms identified key features influencing anemia, which informed a mechanistic model based on ordinary differential equations and mixed effects. Most parameters were fixed globally, but three—ESA binding sites, hemoglobin degradation rate, and initial hemoglobin level—were identified as patient-specific. Altogether, these results provide the basis to develop a decision support tool that aims to stabilize haemoglobin levels in patients, thereby lowering the risk of thromboembolic events, and reducing ESA treatment costs. To support clinical integration, we are currently focusing on real-time data retrieval and analysis. Our main goal is to deliver therapy recommendations within minutes, enabling precision dosing for maintaining erythropoietic homeostasis, and improve the quality of life of CKD and cancer patients. |
|
A-419: NeuralFlux: Estimation of reaction fluxes at a genome-scale level from time-resolved isotope labeling patterns using deep learning.
Intracellular metabolic fluxes provide insights into molecular mechanisms that shape cell physiology, yet they cannot be measured directly. To estimate these fluxes, approaches from Metabolic Flux Analysis (MFA) integrate data on label enrichment of metabolites, from stable isotope labeling experiments, into a model of a reaction network with underlying atom transitions. For experiments with nutrients for which the labeled atoms cannot be differentiated (e.g., carbon in CO2), the stationary state of label enrichment is uninformative and time-resolved data from the transient labeling must be used. The inverse problem solved by MFA approaches is the estimation of both a steady-state flux distribution and (compartmentalized) metabolite concentrations for which the simulated label enrichments minimize a chosen distance metric to the measured enrichments. Given that label enrichment is simulated by solving a system of ordinary differential equations, MFA approaches for genome-scale flux estimation from instationary label enrichment data have prohibitive computational costs. Here, we present NeuralFlux, a deep learning approach that predicts time-resolved labeling patterns with different isotope labels from a steady-state flux distribution and metabolite concentrations, leading to a three-order-of-magnitude decrease in computational cost for flux estimation. In our proof-of-principle, we employ NeuralFlux to estimate fluxes in a large-scale metabolic network of Arabidopsis thaliana using isotope labeled carbon dioxide and/or nitrate. Our results demonstrate that NeuralFlux can effectively estimate fluxes, but also show that expanded coverage of measured metabolites is needed to increase the precision of estimates at a genome scale. |
|
A-421: New methodological approaches for human precision toxicology
Denise Duma, Center for Genomic Regulation, Spain
Anna Vlot, Oak Ridge National Laboratory, United States
Malini Nemalikanti, Center for Genomic Regulation, Spain
Carsten Weiss, Karlshrue Institute of Technology, Germany
Robert Russell , University of Heidelberg, Germany
John Colbourne, University of Birmingham, United Kingdom
Roderic Guigo, Center for Genomic Regulation, Spain
More people die prematurely from environmental exposure to toxic chemicals than from war, hunger, malaria, AIDS, or tuberculosis. The PrecisionTox project aims to discover biomolecular toxicity pathways and biomarkers relevant for the protection of human health by leveraging multi-omics data generated from the exposure of non-sentient model organisms to an array of carefully selected test chemicals. In this study, we report findings from the bulk RNA-seq analysis conducted during the pilot phase of the project. This involved treating five model organisms—Caenorhabditis elegans, Daphnia magna, Drosophila melanogaster, Danio rerio, and Xenopus laevis— as well as human HepG2 cell lines with four selected toxic substances plus the solvent dimethyl sulfoxide (DMSO). Each substance was applied in replicate at two distinct doses, and samples were collected at three to five time points, depending on the species. The transcriptome data generated from these samples underwent rigorous quality control and bioinformatic processing to identify functionally conserved differentially expressed genes and pathways among the distantly related species. Subsequently, we developed a bioinformatics pipeline designed to validate the known biomolecular effects of these chemicals as well as reveal novel toxicological effects and potential mechanisms of action. Our results demonstrate that the end-to-end bioinformatic approach we implemented can robustly identify both species-specific and evolutionarily conserved toxicological responses, providing new insights into shared mechanisms of toxicity across diverse biological systems. |
|
A-423: Dynamic modeling of the transcriptional regulatory network controlling cellular senescence
José-Américo Freitas, Mondor Institute for Biomedical Research (IMRB), France
Oliver Bischof, Mondor Institute for Biomedical Research (IMRB), France
Dynamic systems theory has increased our predictive power in a wide variety of industrial and healthcare settings, with notable applications in the robust control of complex systems. Interestingly, control strategies are implemented at the molecular level, notably by gene regulatory networks (GRN). This equips cellular systems to maintain balance despite minor disturbances. However, larger molecular damage can cause significant and lasting changes in cell characteristics in a process known as cellular senescence (CS). Chronic CS increase is known to be associated with various age-related diseases, such as diabetes, cancer, and Alzheimer’s disease. In order to understand the GRN that implements changes in the cell’s state and identity, we leverage the comprehensive scRNA-seq atlas Tabula Sapiens and RNA velocity inference tools to build a non-linear dynamic system that mimics the transcriptome evolution of senescent cells. We employ model interpretability tools, such as Local Interpretable Model-Agnostic Explanations (LIME), to inform us about the regulators with highest weights for a given target. By characterizing non-linear interactions, our model accounts for regulatory interactions dependent on cell identity and is able to provide predictions tailored to cell type, using the same set of equations. Ultimately, we will be able to quantify cellular senescence dynamics and apply control theory principles to restore homeostasis, identifying actionable targets with therapeutic applications and alleviating the detrimental impact of senescent cells on aging and age-related diseases to prolong healthspan. |
|
A-425: Toward Frustration-Free Proteins: A Multiobjective Strategy for Navigating Evolutionary Trade-offs
Martin Floor, Barcelona Supercomputing Center, Spain
Albert Cañellas-Solé, Barcelona Supercomputing Center, Spain
Miriam Poley-Gil, Barcelona Supercomputing Center, Spain
R. Gonzalo Parra, Barcelona Supercomputing Center, Spain
Victor Guallar, Barcelona Supercomputing Center, Spain
Introduction Although the principle of minimal frustration guides proteins toward stable conformations, regions of energetic conflict persist in most structures. These “frustrated” regions are often correlated to function yet can conflict with stability and designability. A general question to investigate is whether frustration can be reduced beyond the limit of natural proteins (~10%). Here we present a multiobjective protein design framework that balances competing constraints: decreasing local frustration, preserving structural integrity, and maintaining biophysical plausibility. Methods Our approach combines a genetic algorithm with three objectives: Rosetta energy for structural stability, a Frustratomer-based score to identify and minimize energetic conflicts, and a ProteinMPNN-derived sequence likelihood to prioritize sequences consistent with patterns learned from natural proteins. By tuning these contributions, we navigate sequence space, identify variants with reduced energetic conflicts, and broaden our understanding of design constraints. We have tested our method on different natural and de novo designed proteins. Results Our methodology provides a platform for exploring how frustration influences protein function, stability, and design feasibility, offering an alternative lens to natural evolutionary processes. We discuss methodological implications and how such designs can be assessed through computational simulations. Our approach lays the groundwork for rethinking protein design in light of the dynamic compromises encoded by nature and invites further inquiry into whether minimizing frustration is compatible with protein function. |
|
A-427: Analysis of Ligand Diffusion Dynamics in Tissue Microenvironments Using High-Resolution Spatial Omics
Haruka Hirose, Laboratory of Computational Life Science, National Cancer Center Research Institute, Japan
Yasuhiro Kojima, Laboratory of Computational Life Science, National Cancer Center Research Institute, Japan
Cell-cell interactions are imperative for the maintenance of tissue homeostasis. Dysregulation of these interactions has been linked to the development of various diseases, including cancer and autoimmune conditions. Consequently, a comprehensive understanding of their mechanisms is imperative for the development of novel therapeutic interventions. Ligand proteins, which have been identified as key signalling molecules in cell-cell interactions, bind to cell surface receptors and trigger intracellular signalling pathways. While the interactions between neighbouring cells have been empirically validated, the spatial range of secreted ligands remains to be comprehensively investigated. In this study, we initially modelled ligand spatial diffusion using Visium data. Assuming a Gaussian distribution, a model was developed to illustrate the diffusion of ligands and the expression of target genes. The objective variable was defined as the expression of target genes, and the action range of each ligand was estimated by determining the diffusion distance as a parameter. This study was conducted utilising high-resolution StereoSeq data, a development that facilitated single-cell-level analysis and represented a significant advancement in the field. Single-cell analysis has the potential to offer a more comprehensive understanding of the spatial dynamics of intercellular signalling and the influence of the microenvironment on signalling pathways. These findings may lead to a better understanding of the mechanisms of diffusible ligand-mediated intercellular signalling and to more targeted therapeutic strategies to modulate specific cell-cell interactions in disease. |
|
A-429: LPD: a synthetic dataset of ladybird wing color patterns generated through extensive numerical analysis of a reaction-diffusion PDE model
We present the Ladybird Pattern Dataset (LPD), a synthetic dataset of ladybird wing color patterns generated using a reaction-diffusion PDE model. LPD contains 70,000 images categorized into 10 classes, with each data entity comprising the kinetic parameters and initial conditions of the PDE model, along with the resulting image of ladybird pattern. LPD serves as a benchmark dataset for classification and generative modeling tasks, similar to MNIST or FashionMNIST. In particular, LPD provides both dynamic simulation parameters and visual patterns, making it suitable as a benchmark for physics-informed and interpretable generative models. Unlike conventional image datasets in machine learning, LPD enables researchers to analyze how biological mechanisms drive developmental pattern formation, specifically inspired by the pattern formation of ladybirds. LPD will be publicly available to facilitate research in computational biology, biological modeling, and AI. |
|
A-431: Computational Modeling of Shortening and Reconstruction of Telomeres Marie Doumic, INRIA, France
We discuss a stochastic model of growth of a cell population of cultured yeast cells with gradually decaying chromosome endings called the telomeres, as well as models of telomere reconstruction using the so-called ALT (alternate lengthening of telomeres) Mechanism. Telomeres play a major role in aging and carcinogenesis in humans. Our models correspond in part to the experiments of one of us (Teixeira). For telomere shortening, we modify the method of Olofsson and Kimmel, who considered properties of the branching process of telomere shortening; this leads to consideration of a random walk on a two-dimensional grid. We derive an integral equation for the probability generating functions (pgf’s) characterizing the dynamics of shortening of telomeres. We find that the general solutions have the form of exponential polynomials. Stochastic simulations lead to interesting and non-obvious effects if cell death is included. We further consider more complex models, involving cell death and the ALT mechanism. of telomere reconstruction. These are based on our works and are intended to address the experiments in Kockler et al. (2021). In one version of the ALT Mechanism model, we consider expectations conditional on non-extinction, since only a fraction of ALT telomeres is stably elongated (see Kockler et al. 2021). As a conclusion, the multitype branching processes produce realistic prediction concurrent with complex biological experiments involving telomeres. We aim to use the models for longer-term prognoses for human telomeres. |
10:00-11:20 and 16:00-16:40 | Session B: July 22, 2025 |
---|---|
B-390: Spatialproteomics – an interoperable toolbox for analyzing highly multiplexed fluorescence image data
Matthias Meyer-Bender, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany, Germany
Harald Voehringer, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany, Germany
Christina Schniederjohann, University Hospital Düsseldorf, Düsseldorf, Germany, Germany
Peter-Martin Bruch, University Hospital Düsseldorf, Düsseldorf, Germany, Germany
Sarah Koziel, University Hospital Düsseldorf, Düsseldorf, Germany, Germany
Erin Chung, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany, Germany
Sascha Dietrich, University Hospital Düsseldorf, Düsseldorf, Germany, Germany
Wolfgang Huber, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany, Germany
Highly multiplexed immunofluorescence imaging is a recent method to characterize tissues at single-cell resolution on the protein level, offering low cost, high scalability, and the ability to analyze paraffin-embedded tissue samples. However, the analysis of these data involves a sequence of steps, including segmentation, image processing, marker quantification, cell type classification, and neighborhood analysis, each of which involves a multitude of method and parameter choices that need to be adapted to the data and analytical objective at hand. Moreover, variations in data quality can be high and unpredictable, which necessitates further flexibility and interactivity. While individual components exist, there is an unmet need for a coherent toolbox that offers end-to-end coverage of the workflow, flexibility, and automation.
We present spatialproteomics, a python package that addresses these challenges. Built on top of xarray and dask, spatialproteomics can process images that are larger than the working memory. It supports synchronization of shared coordinates across data modalities such as images, segmentation masks, and expression matrices, which facilitates easy and safe subsetting and transformation. We demonstrate spatialproteomics on a set of images of reactive lymph nodes or different forms of B cell Non-Hodgkin lymphomas (BNHL) from 132 patients. We showcase an end-to-end analysis from raw images to statistical characterization of cell type composition and spatial distribution across indolent and aggressive lymphomas. Furthermore, we show how spatialproteomics can process gigapixel whole slide images. Altogether, we propose spatialproteomics as an easy-to-install, easy-to-learn, comprehensive toolbox for constructing powerful end-to-end image analysis solutions for highly multiplexed immunofluorescence imaging. |
|
B-392: Systems-Level Analysis of Metabolic Dysregulation in Gaucher Disease: Mitochondrial Dysfunction and Disrupted Cholesterol Homeostasis
Yanjun Liu, University of Galway, Ireland, Ireland
Xi Luo, University of Galway, Ireland, Ireland
Samira Ranjbar, University of Galway, Ireland, Ireland
Ronan Fleming, University of Galway, Ireland, Ireland
Gaucher disease (GD), a lysosomal storage disorder caused by GBA1 mutations, is characterized by glucosylceramide accumulation and profound metabolic alterations in macrophages. To investigate these alterations at a systems level, we developed personalized genome-scale metabolic models of GD and healthy macrophages by integrating transcriptomic and exometabolomic data. Model simulations revealed extensive metabolic reprogramming in GD, including a shift from oxidative phosphorylation to glycolysis under energy stress, consistent with impaired mitochondrial function and reduced ATP output. Differential flux and correlation analyses further uncovered dysregulated amino acid metabolism, characterized by elevated uptake and nitrogen waste secretion, suggestive of enhanced catabolism and mitochondrial stress. Beyond primary sphingolipid accumulation, we identified widespread lipidomic disruptions, including impaired mitochondrial β-oxidation and altered fatty acid handling. Notably, cholesterol metabolism consistently emerged as a central axis of dysregulation across all analyses, involving changes in flux, transcriptional control, and co-regulated metabolic modules. These findings align with clinical lipid profile abnormalities in GD and highlight cholesterol homeostasis as a potential therapeutic target. Our study underscores the utility of constraint-based modeling in elucidating complex metabolic phenotypes and offers mechanistic insights into the pathophysiology of Gaucher macrophages. |
|
B-394: KINEPIK: An integrated data resource for cell signalling research
Venla Heinonen, Queen Mary, University of London, United Kingdom
Magdalena Hübner, Queen Mary, University of London, United Kingdom
Pedro Cutillas, Queen Mary, University of London, United Kingdom
Conrad Bessant, Queen Mary, University of London, United Kingdom
Kinase signalling plays a vital role in many biological processes, including the initiation and progression of disease. In recent years, phosphoproteomics has become an indispensable tool for exploring how kinase activity changes in response to perturbations. By interpreting phosphoproteomics data from perturbed cells in the context of existing knowledge (e.g. known kinase-substrate relationships) researchers are uncovering new insights into cell signalling pathways. However, bringing together the disparate data needed for this type of analysis is a hinderance to such work because the data is distributed across different sources, using inconsistent molecular nomenclature. To address this problem, we present KINEPIK, an integrated data resource designed to streamline cell signalling research. At its core, KINEPIK features a relational database that links kinases with their known phosphosite targets, perturbations that influence their activity, and other information such as subcellular location. This database is integrated with a robust framework for storing high-quality phosphoproteomic data generated from studies involving genetic or chemical perturbations. Although initially focused on human data, KINEPIK is designed to support multiple species. Developed using an API-first approach, KINEPIK offers seamless access to its data from popular environments such as Python and R, enabling smooth integration into analytical workflows. We demonstrate the platform’s capabilities through two common use cases: quantifying kinase activity from phosphoproteomic profiles, and building a cell signalling knowledge graph. |
|
B-396: An informed deep learning model of the Omicron wave and the impact of vaccination
Elham Shamsara, Methods in Medical Informatics, Department of Computer Science, University of T¨ubingen, 72076 T¨ubingen, Germany, Germany
Florian König, Methods in Medical Informatics, Department of Computer Science, University of T¨ubingen, 72076 T¨ubingen, Germany, Germany
Nico Pfeifer, Methods in Medical Informatics, Department of Computer Science, University of T¨ubingen, 72076 T¨ubingen, Germany, Germany
The Omicron (B.1.1.529) variant of SARS-CoV-2 emerged in November 2021 and has since evolved into multiple lineages. Understanding its transmission, vaccine efficacy, and potential for reinfection is crucial. This study examines the dynamics of Omicron in Germany, France, and Italy by employing Physics-Informed Neural Networks to estimate the temporal parameters influencing its spread. We validated the performance of our model using the Root Mean Squared Percent Error (RMSPE). Link to the paper: https://www.sciencedirect.com/science/article/pii/S0010482525003191 |
|
B-398: Advancing diagnostics in inherited metabolic diseases through enhanced metabolic modeling of lipid metabolism
Anastasia Sveshnikova, SIB Swiss Institute of Bioinformatics, Switzerland Lucila Aimo, SIB Swiss Institute of Bioinformatics, Switzerland
Jerven Bolleman, SIB Swiss Institute of Bioinformatics, Switzerland
Sebastien Moretti, SIB Swiss Institute of Bioinformatics, Switzerland
Anne Niknejad, SIB Swiss Institute of Bioinformatics, Switzerland
Marco Pagni, SIB Swiss Institute of Bioinformatics, Switzerland
Nicole Redaschi, SIB Swiss Institute of Bioinformatics, Switzerland
Alan Bridge, SIB Swiss Institute of Bioinformatics, Switzerland
Inherited metabolic diseases (IMDs) often manifest with diverse, non-specific symptoms that resemble more common conditions, leading to delayed diagnoses and reduced patient quality of life. The Horizon Europe project Recon4IMD seeks to improve IMD diagnosis by advancing computational metabolic modeling. A central effort is the development of Recon4, a human metabolic model integrating Recon3 with key resources such as the Virtual Metabolic Human (VMH), MetaNetX, UniProtKB/Swiss-Prot, and the Rhea reaction database. Our contribution focuses on chemically and catalytically precise representation of human lipid metabolism. Using SwissLipids and Rhea, we systematically enumerated 2,253,663 unique lipid-specific reactions with RInChIKeys, formatted for integration into MetaNetX. To support this, we developed the Python package pyrheadb, which translates Rhea into a usable network and model, with documentation provided at GitHub. To address disconnected compounds, 1,849 reaction templates from Rhea generated 1,350 candidate reactions aimed at closing 62 network gaps. Additionally, we extended reaction enumeration to stereospecific variants by detecting undefined stereocenters, generating stereochemical forms, and applying atom mapping, resulting in stereo options for 2,042 Rhea reactions. Sphingolipid reactions posed unique challenges due to their glycan-based nomenclature. Using SphinGOMAP and curator-supplied data, we translated sugar chain formulas into reaction SMILES, producing 2,000 balanced reactions for model integration. This work enhances lipid metabolic pathway reconstructions with high chemical specificity, supports the transition to Recon4, and introduces database-independent identifiers to improve reproducibility. It also lays the groundwork for IMD-specific knowledge graphs and sex-specific metabolic models. |
|
B-400: SPECIMEN and refineGEMs: Open-Source Python Tools for Streamlined Metabolic Modelling
Gwendolyn Olivia Döbel, Martin Luther University Halle-Wittenberg, Germany
Carolin Brune, Martin Luther University Halle-Wittenberg, Germany
Andreas Dräger, Martin Luther University Halle-Wittenberg, Germany
Systems biology aims to understand organisms by analysing them as context-specific However, building high-quality models is time-consuming, as many steps require manual input. The open-source Python toolbox refineGEMs provides a variety of functionalities for model curation and a novel, in silico media database for simulating models under different conditions. Building upon this toolbox, SPECIMEN is an open-source Python package containing a collection of primarily automated workflows for model curation. For example, it includes a workflow based on draft model generation via CarveMe (CMPB) and another, where the draft model is generated using a template model and performing a bidirectional BLAST (HQTB). In addition to the initial model generation, the workflows include further steps, including gap-filling (identifying missing reactions in pathways and adding them to the model), charge curation (balancing the charges in the chemical reactions), biomass objective function (chemical equation describing biomass production) adjustment, the addition of annotations from various databases, and more. Overall, the workflows allow for the efficient generation of new models of higher quality and completeness than the outputs of standard workflows for draft model generation, making constrained-based modelling faster and more accessible. |
|
B-402: A Co-Evolutionary Multi-Objective Framework for Joint Topology and Parameter Discovery in Gene Regulatory Networks
Tuan Do, N2TP Technology Solutions JSC., Viet Nam Nhung Duong, Hanoi University of Pharmacy, Viet Nam
Tien Nguyen, Hanoi University of Pharmacy, Viet Nam
Ngoc Do, Hanoi University of Pharmacy, Viet Nam
Anh Truong, Hanoi University of Pharmacy, Viet Nam
Lap Nguyen, Hanoi University of Pharmacy, Viet Nam
Background: Reconstructing Gene Regulatory Networks (GRNs) requires inferring both network topology and kinetic parameters, a persistent challenge in systems biology often tackled separately. Biological realism, however, demands simultaneous optimization under multiple constraints. We aimed to develop a framework for joint topology and parameter discovery that explicitly balances dynamic accuracy, network simplicity, and system robustness.
Methods: We present a co-evolutionary multi-objective optimization framework using two interacting populations for topologies and ODE-based kinetic parameter sets (Hill-type models). The algorithm simultaneously optimizes three objectives: (1) minimizing Mean Squared Error (MSE) against target time-series data, (2) promoting network parsimony (fewer edges), and (3) maximizing robustness to perturbations. Biologically-inspired evolutionary guide the search. Results: On synthetic GRN benchmarks (N=5 and N=10 genes), our co-evolutionary approach significantly outperformed baseline methods in capturing target dynamics. For instance, it achieved lower MSE than single-objective optimization (e.g., 0.136 vs 0.174 for N=5; 2.172 vs 2.792 for N=10) and often surpassed even methods given the true topology. While exact structural recovery was partial (F1 scores 0.2-0.6 vs. 1 for topology-fixed), the framework consistently identified functionally equivalent networks that accurately reproduced dynamics using alternative, often parsimonious, structures while maintaining high robustness. Conclusion: Our co-evolutionary multi-objective framework effectively navigates the complex search space of joint GRN inference. By prioritizing functional equivalence and balancing competing objectives, it discovers dynamically accurate and robust network models, even without perfect structural matches. This provides a valuable tool for systems biology analysis and synthetic biol |
|
B-404: Spatial Flux Balance Analysis reveals region-specific cancer metabolic rewiring and metastatic mimicking
Davide Maspero, Centro Nacional de Análisis Genómico, Barcelona, Spain, Spain
Giovanni Marteletto, Immunobiology Department, Yale School of Medicine, New Haven, USA, United States
Francesco Lapi, Department of Biosciences and Biotecnologies, University of Milano-Bicocca, Milan, IT, Italy
Bruno Giovanni Galuzzi, Institute of bioimaging and complex biological systems (IBSBC), Segrate, IT, Italy
Irene Ruano, Centro Nacional de Análisis Genómico, Barcelona, Spain, Spain
Ben Vandenbosch, Department of Oncology, KU Leuven, Leuven, Belgium, Belgium
Ke Yin, Department of Oncology, KU Leuven, Leuven, Belgium, Belgium
Sabine Tejpar, Department of Oncology, KU Leuven, Leuven, Belgium, Spain
Alex Graudenzi, Department of Informatics, Systems and Communication, University of Milano-Bicocca, Milan, IT, Italy
Holger Heyn, Universitat de Barcelona (UB), Barcelona, Spain, Spain
Anna Pascual-Reguan, Centro Nacional de Análisis Genómico, Barcelona, Spain, Spain
Chiara Damiani, Department of Biosciences and Biotecnologies, University of Milano-Bicocca, Milan, IT, Italy
To investigate cancer cell metabolism with spatial precision, it is crucial to estimate the metabolic fluxes with spatial resolution. However, current methods for inferring metabolic fluxes lack spatial resolution at the scale now achievable with spatial transcriptomics.
To address this gap, we introduce spatial Flux Balance Analysis (spFBA), a novel framework to process spatial transcriptomics data and compute metabolic fluxes at the spatial spot resolution. The spFBA approach builds upon previous work designed for bulk and single-cell data to return metabolic fluxes, up to the level of single reactions, that can distinguish their preferred directional usage, retaining spatial resolution. spFBA integrates differential constraints on flux boundaries in steady-state metabolic modelling, informed by spatial gene expression, and utilizes a corner-based sampling strategy. We validated spFBA using a publicly available renal cancer dataset, including tumor-normal interface samples. spFBA successfully summarises histological structures and detected cancer metabolic hallmarks, such as enhanced glucose uptake, lactate production, and metabolic growth. Yet, it achieved unprecedented resolution, uncovering region-specific features, including increased glutamate consumption at the tumor interface and hypoxic regions within the tumor core. By applying spFBA to a new stereo-seq colorectal cancer dataset, including paired primary tumor and liver metastasis samples, we provide compelling evidence that metastases mimic the metabolic traits of their tissue of origin. Additionally, we present the first in vivo evidence of lactate-consuming cancer cells. spFBA stands out as a powerful framework to unravel the spatial metabolic complexity of cancer and beyond, leveraging the expanding landscape of spatial transcriptomics datasets. |
|
B-406: Modeling differentiation at clonal level in B-cell Precursor Acute Lymphoblastic Leukemia (BCP-ALL)
Adrien Jolly, University Hospital Frankfurt, Germany
Alec Gessner, University Hospital Frankfurt, Germany
Michael Rieger, University Hospital Frankfurt, Germany
Florian Buettner, University Hospital Frankfurt and German Cancer Research Center, Germany
Adult BCP-ALL is a neoplastic disease characterized by phenotypical and functional heterogeneity. However, the cellular organization underlying this heterogeneity is poorly understood. It can arise from both clonal evolution and/or from cell differentiation but so far no result could conclusively determine whether differentiation occurs. To address this question, we use BCP-ALL patient-derived long-term cultures. We have developed expressible barcode vectors that can be detected alongside the transcriptomes and surface markers following lentiviral transduction. The barcoded cells are transplanted in immuno compromised ,mice to study leukemogenesis in vivo. We performed single-cell CITE-seq at full-blown leukemia which revealed that a majority of clones spanned multiple cell populations, indicating differentiation between phenotypically distinct cell clusters. |
|
B-408: Cancer cell-intrinsic programs of alternative splicing in tumour development and progression
Larisa M. Soto, Department of Human Genetics, McGill University, Canada
Hamed S. Najafabadi, Department of Human Genetics, McGill University, Canada
Widespread alternative splicing (AS) alterations have been reported in tumour transcriptomes of all known cancer types, but it is unclear to what extent these alterations reflect cancer cell-intrinsic reprogramming of the splicing landscape as opposed to variations in tumour cellular composition—tumour infiltrating cells have vastly different splicing landscapes, potentially creating a large number of false positive splicing changes. This and other confounding factors have been largely neglected in previous studies, despite the compelling evidence supporting the impact of confounder biases in RNA sequencing analyses.
Using TRex, a novel computational tool to model the impact of multiple variables on the quantification of differential AS from RNA-seq data, we estimated changes between (1) tumour and normal tissues and (2) tumours of different stages using 8,633 samples spanning 24 cancer types. Our systematic analysis revealed that tumour purity explains up to 75% of the variance in splicing profiles between tumours of the same cancer type. Intron retention (RI) emerges as the most common cancer cell-intrinsic program, with some RI events recurrently dysregulated in nearly two-thirds of the cancers. Myogenesis, Mitotic Spindle, and other cancer hallmark pathways are strongly enriched among cancer-cell AS events dysregulated in multiple cancer types, whereas immune-related pathways are overrepresented in impurity-associated splicing changes. This work presents a comprehensive, unbiased compendium of cancer cell-intrinsic AS programs contributing to different axes of tumour biology. This resource opens the door to studying mechanisms of AS reprogramming in cancer cells and their contribution to disease development and trajectory. |
|
B-410: Inferring Trajectories of Cell State Transition with Real-Time Scaling
Kaitlyn Ramesh, Northeastern University, United States
Benjamin Clauss, Tufts University, United States
Mingyang Lu, Northeastern University, United States
Cell phenotypic transitions are influenced by dynamic changes in gene expression over time. To investigate these dynamics using static single-cell gene expression data, pseudotime algorithms have been developed. However, existing methods often fail to recapitulate a biologically accurate timescale. Here, we present RETRO, a pseudotime algorithm that integrates gene expression data and low-resolution temporal information to accurately infer transcriptional dynamics. We applied RETRO to both synthetic and experimental datasets. Compared to existing methods, RETRO demonstrates improved performance in identifying complex trajectories and timescales of cell state transitions. Further application of RETRO will enable deeper mechanistic insights into cell fate decision.
|
|
B-412: Metabolic Modeling of Anaerobic Digestion Process in Waste Activated Sludge by Transcriptomic Data Integration
Ebru Sarsılmaz, Bezmialem Vakıf University, Turkey
Muhammed Erkan Karabekmez, Istanbul Medeniyet University, Turkey
Anaerobic digestion of waste activated sludge represents a key biotechnological strategy for the sustainable treatment of organic waste and the recovery of bioenergy, particularly in the form of methane. This complex process is mediated by a consortium of microorganisms that coordinate the breakdown of organic matter through syntrophic interactions. Understanding the metabolic activity of individual microbial taxa within this community is essential for optimizing process efficiency and stability. In this study, we integrate transcriptomic data with genome-scale metabolic models (GEMs) to investigate the metabolic capabilities and interactions of five key microbial genera involved in anaerobic digestion. Due to inconsistencies between transcriptomic gene identifiers and those used in available metabolic models, a gene ID harmonization step was implemented to enable accurate integration. GEMs were curated and refined by incorporating expression data, removing non-essential reactions, and downscaling reactions with low transcriptional support. Flux Balance Analysis (FBA) was subsequently performed to estimate organism-specific growth rates under anaerobic conditions. Beyond individual modeling, these curated GEMs are being integrated into a community-scale model to explore interspecies metabolic dependencies and syntrophic interactions. Although results are pending, this systems-level approach is expected to provide mechanistic insights into the metabolic architecture of anaerobic sludge communities and inform strategies to enhance biogas yield and process resilience. |
|
B-414: Transcriptomic landscape of Olaparib treatment in high-grade serous ovarian cancer
Roxana Andreea Moldovan, Department of Applied Statistics, Operations Research and Quality, Universitat Politècnica de València, Spain, Spain
Chiara Battistini, Unit of Gynaecological Oncology Research, European Institute of Oncology IRCSS, Milan, Italy, Italy
Ugo Cavallaro, Unit of Gynaecological Oncology Research, European Institute of Oncology IRCSS, Milan, Italy, Italy
Mirjana Kessler, Department of Obstetrics and Gynaecology , LMU University Hospital, Munich, Germany, Germany
Sophia Geweniger, Department of Obstetrics and Gynaecology , LMU University Hospital, Munich, Germany, Germany
Sonia Tarazona, Department of Applied Statistics, Operations Research and Quality, Universitat Politècnica de València, Spain, Spain
High-grade serous ovarian cancer (HGSOC) is the most prevalent and aggressive form of gynaecological cancer, accounting for approximately 75% of cases. Despite advancements in therapeutic interventions, there has been no substantial improvement in overall patient survival over the past few decades. A therapeutic strategy that has been employed involves the inhibition of PARP (Poly ADP-Ribose Polymerase), a key enzyme in DNA repair. Drugs such as Olaparib, a PARP inhibitor, have demonstrated efficacy in patients with BRCA1/2 mutations and/or homologous recombination deficiencies. However, the molecular mechanisms underlying this response remain to be fully elucidated. The present study aims to elucidate the molecular mechanisms underlying the response and resistance to Olaparib treatment by analysing transcriptional changes using in vitro models derived from HGSOC patients. Samples were treated with DMSO or Olaparib and subsequently analysed using RNA-Seq. The bioinformatic processing comprised the following steps: quality control (FastQC, FastP), alignment to the reference genome (STAR), quantification (FeatureCounts), and analysis in R, including filtering (NoiSeq), normalization (Conditional Quantile Normalization), batch effect correction (MultiBaC), differential expression analysis (limma), and functional analysis (GSEA, ORA). The results revealed alterations in key genes, highlighting a decrease in ASPM expression, which is involved in DNA repair and cell proliferation. Furthermore, at functional level, the study identified a significant alteration regarding the transcriptional regulation of RUNX, a critical transcription factor in HGSOC. These findings contribute to the understanding of PARP inhibition mechanism and could offer new perspectives in the search for biomarkers and therapeutic strategies in HGSOC. |
|
B-416: Computational Prediction of Targeted Nutritional Interventions for Host–Microbiome Modulation in C. elegans
Rupinder Kaur, UKSH Campus Kiel, Germany
Prof. Dr. Christoph Kaleta, UKSH Kiel, Germany
Rational dietary modulation of host-associated microbiota demands approaches that translate multi-omics data into mechanistic, testable hypotheses. We present a computational framework combining dual RNA-sequencing, genome-scale metabolic modelling, and spatially explicit agent-based simulation to prioritise nutritional interventions in the Caenorhabditis elegans gut ecosystem. Wild-type C. elegans were exposed to eight bacterial strains, including pathogens (Pseudomonas aeruginosa, Enterococcus faecalis) and commensals (Sphingobacterium sp., Ochrobactrum sp., and four additional isolates). Paired-end reads are analysed through both co-mapping (concatenated reference) and sequential host-to-microbe alignments; resulting gene-level counts constrain condition-specific genome-scale metabolic models (GEMs) for C. elegans and each bacterium. Flux-consistent pruning removes thermodynamically infeasible loops, yielding context-specific metabolic networks. An automated nutrient-scanner systematically perturbs single or paired dietary substrates, and flux balance analysis predicts effects on bacterial biomass, cross-feeding interactions, and proxies for host fitness. To incorporate spatial heterogeneity, GEMs are embedded into WormColon, an agent-based spatial module adapted from BacArena’s virtual-colon concept, simulating microbial colonisation and nutrient diffusion within the gut lumen. Current milestones include completion of all condition-specific GEMs and the nutrient-scanner implementation; spatial simulations are ongoing. The framework outputs a ranked catalogue of candidate supplements annotated with mechanistic reaction signatures and spatial growth dynamics. Experimental validation will combine controlled feeding assays, follow-up dual RNA-seq, and lifespan analyses, enabling iterative model refinement. This approach provides a scalable, mechanistic route from dual-omics data to actionable dietary strategies, advancing precision microbiome engineering in simple animal models. |
|
B-418: Transcriptome Analysis and Genome‑Scale Metabolic Reconstruction of Bos taurus for Optimizing Cultured Meat Production
Junkyu Lee, KAIST, South Korea
Jaehyun Kim, KAIST, South Korea
Sofia Mun, KAIST, Kazakhstan
Hyun Uk Kim, KAIST, South Korea
Minji Kim, DTU, South Korea
Hyun Won Bae, CJ BIO Research Institute, South Korea
Byung Kwon Jung, CJ BIO Research Institute, South Korea
Juyeon Kim, CJ BIO Research Institute, South Korea
Sunggun Lee, CJ BIO Research Institute, South Korea
Cultured meat industrialization has attracted significant attention as a sustainable alternative to conventional meat production. Bos taurus, a key cattle species, is regarded as a promising source for the cultured meat industry. However, the metabolism of B. taurus muscle cells remains relatively unknown in the context of cultured meat. In this study, transcriptome analysis and genome-scale metabolic reconstruction were performed using bovine satellite cells (BSCs) derived from the semimembranosus muscle of Korean Hanwoo cattle. Differential gene expression and gene set enrichment analyses of RNA-seq data revealed pivotal pathways involved in muscle cell proliferation (e.g., ‘Cell cycle’ and ‘RNA polymerase’) and differentiation (e.g., ‘Cytoskeleton in muscle cells’ and ‘Tryptophan metabolism’). Using the Human1 GEM as a template, we constructed the first B. taurus-specific genome-scale metabolic model, named BtaSBML2986, comprising 2,986 genes, 13,278 reactions, and 8,652 metabolites to perform various metabolic simulations. The model’s predictive accuracy was validated using flux balance analysis (FBA) by comparing simulated growth with experimentally measured growth rates, revealing a strong correlation. BtaSBML2986 was subsequently integrated with the RNA-seq data to examine key pathways such as the TCA cycle and glycolysis associated with cell proliferation. This integrated approach offers insights into critical metabolic pathways for cultured meat production, highlighting the potential of BtaSBML2986 to guide future efforts in optimizing culture processes. |
|
B-420: Reconstructing a genome-scale metabolic model of Haemophilus influenzae strain Rd KW20 to mirror growth within the human nasal environment
Reihaneh Mostolizadeh, Bioinformatics and Systems Biology, Justus Liebig University Giessen, 35392 Giessen, Germany, Germany Niklas Henle, Department of Computer Science, Eberhard Karl University of Tübingen, 72076 Tübingen, Germany, Germany
Oskar Wacker, Department of Computer Science, Eberhard Karl University of Tübingen, 72076 Tübingen, Germany, Germany
Andreas Dräger, Data Analytics and Bioinformatics, Martin Luther University Halle-Wittenberg, Halle (Saale) 06120, Germany, Germany Background: Haemophilus influenzae is a pathogenic bacterium that can cause various diseases, including respiratory tract infections, bloodstream infections, and meningitis. Most of its strains are opportunistic, i.e., they asymptomatically colonize healthy individuals but can cause severe disease in immunocompromised hosts. While antibiotics represent the first-line therapy against H. influenzae infections, their effectiveness is severely undermined by this organism’s spread and diversification of antibiotic resistance.
Results: Comprehensive and reliable genome-scale metabolic models (GEMs) have demonstrated their ability to facilitate our understanding of many organisms. Thus, such a model could help identify new drug targets against the virulence mechanisms across various H. influenzae strains. Here, we present an updated, high-quality H. influenzae strain Rd KW20 model based on more accurate and expanded gene annotations, several biochemical databases, and findable, accessible, interoperable, reproducible (FAIR) data principles. The updated model is a comprehensive knowledge base that fulfills systems biology standards and is available in Systems Biology Markup Language (SBML) format. To match the conditions of the nasal environment, our model mimics the growth conditions of the human nasal mucosa by simulating in silico growth within a modified version of the synthetic nasal medium (SNM). Besides, to give an overview of all available GEMs of H. influenzae, we also examined all previously published models’ growth capabilities and similarities. Furthermore, the open-source quality-control tool MEMOTE was applied to evaluate standardized metabolic tests. Conclusion: This comparison will play a vital role in the more accurate application of GEMs for drug targeting and modeling interactions among multiple organisms. |
|
B-422: Simulating Tumor Evolution with Functional Stochastic Branching Models: A Framework for Testing Evolutionary Hypotheses
Daniela Volpatto, Department of Computer Science, University of Turin, Italy Simone Pernice, Department of Computer Science, University of Turin, Italy Sandro Gepiro Contaldo, Department of Computer Science, University of Turin, Italy Daniela Conticelli, University of Turin, Candiolo Cancer Institute, FPO-IRCCS, Candiolo, Italy Simona Corso, University of Turin, Candiolo Cancer Institute, FPO-IRCCS, Candiolo, Italy
Silvia Giordano, University of Turin, Candiolo Cancer Institute, FPO-IRCCS, Candiolo, Italy
Marco Beccuti, Department of Computer Science, University of Turin, Italy Francesca Cordero, Department of Computer Science, University of Turin, Italy Roberta Sirovich, Department of Mathematics, University of Turin, Italy
We present a synthetic tumor simulator based on stochastic branching processes that captures the dynamics of clonal evolution under biologically interpretable principles. The model tracks cell counts over time for each emerging subclone, shaped by selective pressures and competitive interactions. Clonal fitness is allowed to emerge through five functional mechanisms: (I) proliferative advantage, (II) genomic instability, (III) ecological interactions, (IV) space expansion beyond standard carrying capacity, and (V) passenger mutations. Clones are defined by their genotype, but behave according to phenotype alone i.e. these functional capabilities that are associated with their mutations, minimizing the number of required parameters without losing subclonal resolution. Carrying capacity constraints and clone-specific competition rules shape non-linear dynamics, generating realistic intratumor heterogeneity. Stochasticity is central, allowing rare but impactful events to steer the evolutionary trajectory beyond deterministic expectations.
The simulation outputs include complete clonal genealogies and can be summarized in synthetic VCFs, enabling comparison with deep-sequencing data. We explored different scenarios deriving from different combinations of functional effects with literature data getting insights on what the explanation behind neutral, linear, punctual and parallel evolution might be. Furthermore, we recreated a synthetic counterpart of an in vitro experiment involving MLH1 knockout in mouse cell lines, tailoring the model to mimic the specific conditions of the biological system. By exploring a range of functional configurations, we identified the combination of selective advantages most likely responsible for the observed tumorigenic behavior. This approach systematically tests evolutionary hypotheses and functional mechanisms in a controlled, fully traceable digital environment. |
|
B-424: A query-driven framework for constructing Boolean networks from disease maps: application to Parkinson’s disease
Adrien Rougny, Luxembourg Centre for Systems Biomedicine, University of Luxembourg, Luxembourg
Ahmed Abdelmonem Hemedan, Luxembourg Centre for Systems Biomedicine, University of Luxembourg, Luxembourg
Marek Ostaszewski, Luxembourg Centre for Systems Biomedicine, University of Luxembourg, Luxembourg
Venkata Satagopam, Luxembourg Centre for Systems Biomedicine, University of Luxembourg, Luxembourg
Complex diseases generally involve several interconnected pathways. The understanding of how their cross-talks lead to specific phenotypes is key to deciphering the molecular basis of diseases. To this end, disease maps encode disease-related pathways following graphical and computational systems biology standards. Diagrams of disease maps may support the construction of dynamical models that are necessary to predict effects of perturbations on the molecular mechanisms they describe. Current methods build dynamical models from single diagrams, and are therefore not well suited to capturing the interconnections between pathways. Moreover, they do not support the fine-grained selection of parts of interest from the input diagrams or the output models, which remains a manual task that is difficult to reproduce. Here, we present a novel query-driven framework for the construction of dynamical models from disease maps that overcome these issues. Specifically, our framework allows users to (i) integrate diagrams of one or more disease maps into a graph database; (ii) query parts of interest from these integrated diagrams; and (iii) transform the output of the query into a Boolean network whose dynamics can then be analyzed using state-of-the-art tools. We apply our framework to the Parkinson’s Disease Map, a comprehensive resource capturing key Parkinson’s disease-specific mechanisms. In particular, we show how our framework may be used to study critical pathways leading to disease phenotypes such as neuronal survival, as well as inter-pathway regulations. Finally, we discuss the advantages and challenges of query-based modeling for transparent and reproducible research. |
|
B-426: Computational Membrane Lipid Structure Inspector (CoMLiSI): a tool to verify lipid structural elements in PDB files
Graeme McDowell, National Research Council of Canada, Canada
Anuradha Surendra, National Research Council of Canada, Canada
Qassim Alkassir, National Research Council of Canada, Canada
Steffany Bennett, University of Ottawa, Canada
Miroslava Čuperlović-Culf, National Research Council of Canada, Canada
Biological lipid membranes provide structure for cells and organelles, and either directly or through interaction or modulation of signalling proteins, drive a number of biological processes in both healthy and diseased states. Computational modeling of membranes allows for calculation of properties such as membrane fluidity or curvature, as well as analysis of protein-lipid interactions. Generation of membrane structures requires that the lipid molecules are individually built and combined within the model yet requires manual curation to ensure structural accuracy is maintained in the model. This type of curation is extremely taxing and error prone. |
|
B-428: Interpretable Markov Models Reveal Distinct Transport Pathways and Bottlenecks in the Nuclear Pore Complex
Roi Eliasian, The Hebrew University of Jerusalem, Israel
Barak Raveh, The Hebrew University of Jerusalem, Israel
The Nuclear Pore Complex (NPC) forms a selective barrier between the nucleus and cytoplasm, efficiently transporting diverse molecular cargos while restricting passive diffusion based on cargo size. Despite its critical biological importance, fundamental questions persist regarding how actively transported molecules navigate the layers of disordered FG repeats lining the NPC’s central channel. To elucidate these transport mechanisms, we developed interpretable Markov models derived from coarse-grained Brownian dynamics simulations, capturing detailed transition dynamics between distinct FG-repeat domains of various FG nucleoporins (FG Nups). Consistent with recent experimental imaging, our analysis reveals transport predominantly occurring through eight spatially distinct conduits, characterized by minimal lateral diffusion between NPC spokes, supporting a parallel-channel transport model. Ongoing analyses examine whether this vertical transport preference remains consistent across varying cargo sizes and transport conditions. By mapping specific transport sequences, we identified characteristic “relay race” patterns, whereby cargos sequentially transfer between critical nucleoporins during transit. Systematic perturbation studies involving targeted deletion of key FG Nups pinpointed essential transport bottlenecks and assessed robustness against structural disruptions. Our Markov-based framework facilitates rapid hypothesis testing of NPC transport dynamics, overcoming computational limitations inherent to direct simulation, and providing crucial insights into how this essential cellular gateway achieves selectivity amid diverse molecular traffic. The suggested approach can describe diverse networks of dynamic protein-protein interactions, including those in biomolecular condensates.
|
|
B-430: Network-centric analysis of a post-traumatic stress disorder regulatory model
Fabien Romano, INSERM, France
Clémence Réda, Universität Rostock, France
Andrée Delahaye-Duriez, INSERM U1141, France
Post-traumatic stress disorder (PTSD) is a complex neuropsychiatric condition involving widespread dysregulation, notably in neuronal and immune pathways. Understanding its molecular mechanisms remains a major challenge due to the absence of mechanistic regulatory models and robust biomarkers of sensitivity or resilience. Network-based approaches offer promising perspectives to characterize disease-specific transcriptional signatures and integrate interactions between relevant functional pathways. First, we identified a subset of 524 marker genes by cross-referencing sources of differentially expressed genes from PTSD patients blood transcriptomes. Second, we enriched this set with whole pathways using curated regulatory interactions from the STRING database. Third, we built a boolean network model on the regulation of this set of genes by leveraging gene knockdown perturbation experiments from the LINCS L1000 database. Our model of PTSD comprises 782 genes and 1,605 interactions. We exploited this network to predict key gene regulators whose perturbation shifted attractor states away from PTSD profiles, reversing pathological expression profiles in silico. Then, we partially validated our network by observing that 45.3% of its edges were experimentally confirmed as physical protein-protein interactions. We identified 31 key regulator genes, including IL6, a pro-inflammatory cytokine consistently upregulated in PTSD and correlated with symptom severity. Functional enrichment analysis on these 31 genes revealed strong associations with inflammation, stress-related and depressive disorder-related pathways. Our work introduces a network model of PTSD and a set of prioritized core genes, based on their expression and predicted impact on the network. This integrative approach lays the groundwork for an improved mechanistic understanding of PTSD. |
Key Dates
April 17, 2025
Abstract submission deadline
May 13, 2025
Abstract acceptance notification
May 15, 2025
Late poster submissions deadline
May 22, 2025
Late poster acceptance notifications
Sunday-Thursday July 20-24, 2025
ISMB/ECCB conference
July 22, 2025
SysMod meeting
More information
For more information, please contact the SysMod coordinators 🔗.