Ljubisa Miskovic
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ljubisa Miskovic has a total of 85 co-authors. Here are the top co-authors by number of shared publications:
- Vassily Hatzimanikatis
Swiss Federal Institute of Technology (EPFL)
38 shared publications - Alireza Karimi
Automatic Control Laboratory, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
14 shared publications D. Bonvin
9 shared publications- Milenko Tokic
Swiss Federal Institute of Technology (EPFL)
7 shared publications - Alexandre Sanfelice Bazanella
Depart- ment of Automation and Energy Data-Driven Control Group Universidade Federal do Rio Grande do Sul Porto Alegre- RS Brazil
5 shared publications Bharath Narayanan
Laboratory of Computational Systems Biotechnology, École Polytechnique Fédérale de Lausanne, Switzerland
5 shared publications
Latest Publications
- Open Access
- 13 November 2026
- Published by Elsevier in Cell Systems
Multi-omics-constrained kinetic models quantify pathway rewiring and predict metabolic vulnerabilities in BRCA1-deficient ovarian cancer
- 13 November 2026
- Published by Elsevier in Cell Systems
Open AccessMulti-omics-constrained kinetic models quantify pathway rewiring and predict metabolic vulnerabilities in BRCA1-deficient ovarian cancer
- Open Access
Scalable biophysical constraints for physiologically consistent metabolic states
Systems biology aims to develop predictive models that connect molecular mechanisms to cellular behavior. Genome-scale metabolic models are among the most widely used frameworks for integrating stoichiometric, thermodynamic, and omics-derived information to predict feasible metabolic phenotypes. However, cellular metabolism operates on timescales governed by enzyme kinetics and by the relationship between metabolic fluxes and metabolite pool sizes. In steady-state metabolic models, this relationship can be expressed in terms of metabolite turnover rates, defined as flux-to-pool-size ratios that quantify how rapidly metabolite pools are renewed. As a result, physiologically consistent steady-state solutions should not only satisfy mass-balance and thermodynamic constraints but also exhibit turnover rates consistent with enzyme-mediated cellular dynamics. Current constraint-based approaches can admit many steady-state flux-concentration states that do not account for turnover rates, resulting in phenotypes incompatible with realistic metabolic dynamics, even when multiple types of data are imposed. Here, we present METEOR-K, an optimization framework that links steady-state metabolic fluxes to metabolite concentrations via turnover rate constraints to identify dynamically plausible flux-concentration reference states. Because these constraints reshape the feasible solution space, we also introduce turnover-rate-aware sampling strategies to efficiently explore the resulting feasible region. We applied METEOR-K to models of increasing scope and scale, including a reduced glycolysis pathway, anaerobic E. coli, and near-genome-scale ovarian cancer models. METEOR-K narrowed the admissible steady-state solution space, reduced uncertainty in feasible flux-concentration states, and improved local dynamic behavior. In nonlinear ODE simulations of bioreactor cultivation and drug-response scenarios, METEOR-K-derived states produced intracellular response times compatible with growth-supporting metabolic operation and perturbation recovery. Overall, these results establish metabolite turnover rates as scalable biophysical constraints that improve the physiological consistency of steady-state metabolic modeling. Because turnover rates encode flux-to-pool-size timescale constraints, METEOR-K moves part of physiological-consistency assessment upstream of kinetic parameterization, yielding better-suited flux-concentration reference states for kinetic modeling and dynamic prediction.
Open AccessScalable biophysical constraints for physiologically consistent metabolic states
Systems biology aims to develop predictive models that connect molecular mechanisms to cellular behavior. Genome-scale metabolic models are among the most widely used frameworks for integrating stoichiometric, thermodynamic, and omics-derived information to predict feasible metabolic phenotypes. However, cellular metabolism operates on timescales governed by enzyme kinetics and by the relationship between metabolic fluxes and metabolite pool sizes. In steady-state metabolic models, this relationship can be expressed in terms of metabolite turnover rates, defined as flux-to-pool-size ratios that quantify how rapidly metabolite pools are renewed. As a result, physiologically consistent steady-state solutions should not only satisfy mass-balance and thermodynamic constraints but also exhibit turnover rates consistent with enzyme-mediated cellular dynamics. Current constraint-based approaches can admit many steady-state flux-concentration states that do not account for turnover rates, resulting in phenotypes incompatible with realistic metabolic dynamics, even when multiple types of data are imposed. Here, we present METEOR-K, an optimization framework that links steady-state metabolic fluxes to metabolite concentrations via turnover rate constraints to identify dynamically plausible flux-concentration reference states. Because these constraints reshape the feasible solution space, we also introduce turnover-rate-aware sampling strategies to efficiently explore the resulting feasible region. We applied METEOR-K to models of increasing scope and scale, including a reduced glycolysis pathway, anaerobic E. coli, and near-genome-scale ovarian cancer models. METEOR-K narrowed the admissible steady-state solution space, reduced uncertainty in feasible flux-concentration states, and improved local dynamic behavior. In nonlinear ODE simulations of bioreactor cultivation and drug-response scenarios, METEOR-K-derived states produced intracellular response times compatible with growth-supporting metabolic operation and perturbation recovery. Overall, these results establish metabolite turnover rates as scalable biophysical constraints that improve the physiological consistency of steady-state metabolic modeling. Because turnover rates encode flux-to-pool-size timescale constraints, METEOR-K moves part of physiological-consistency assessment upstream of kinetic parameterization, yielding better-suited flux-concentration reference states for kinetic modeling and dynamic prediction.
- Open Access
- 20 April 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Generative approaches to kinetic parameter inference in metabolic networks via latent space exploration
Dynamic (kinetic) models track time-varying metabolite concentrations, fluxes, and enzyme levels, quantifying responses to genetic and environmental perturbations. Yet building these models at scale is hindered by scarce enzyme kinetic parameters. Generative neural networks can rapidly parameterize near-genome-scale kinetic models, but their representations are hard to interpret and often require new training to move across species or physiological states. Here we introduce a latent-space exploration framework that repurposes a trained generative network to produce models with targeted dynamics in new regimes without additional training. We show in Escherichia coli that latent inputs tune aerobic response speed, identify rate-limiting enzymes, and retarget the generative network to anaerobic dynamics. We extend our approach to Saccharomyces cerevisiae , demonstrating robust control of metabolic dynamics across training stages and diverse latent inputs. Latent variables thus become practical control knobs for kinetic model behavior, accelerating cell-factory design and enabling personalized metabolic modeling.
- 20 April 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessGenerative approaches to kinetic parameter inference in metabolic networks via latent space exploration
Dynamic (kinetic) models track time-varying metabolite concentrations, fluxes, and enzyme levels, quantifying responses to genetic and environmental perturbations. Yet building these models at scale is hindered by scarce enzyme kinetic parameters. Generative neural networks can rapidly parameterize near-genome-scale kinetic models, but their representations are hard to interpret and often require new training to move across species or physiological states. Here we introduce a latent-space exploration framework that repurposes a trained generative network to produce models with targeted dynamics in new regimes without additional training. We show in Escherichia coli that latent inputs tune aerobic response speed, identify rate-limiting enzymes, and retarget the generative network to anaerobic dynamics. We extend our approach to Saccharomyces cerevisiae , demonstrating robust control of metabolic dynamics across training stages and diverse latent inputs. Latent variables thus become practical control knobs for kinetic model behavior, accelerating cell-factory design and enabling personalized metabolic modeling.
- Open Access
Multi-omics-driven kinetic modeling reveals metabolic adaptations and vulnerabilities in BRCA1-deficient ovarian cancer
BRCA1 -deficient ovarian cancer cells undergo extensive metabolic reprogramming, yet the network-level dynamics underlying their proliferation and treatment response remain poorly resolved. Here, we construct large-scale multi-omics-driven kinetic model populations of ovarian cancer metabolism to track how tumor cells adapt to changes in nutrient use, energy production, and metabolite dynamics over time. Across BRCA1 wild-type and mutant cells, these models expose distinct metabolic strategies shaped by transcriptional regulation and prioritize 28 enzyme-mediated vulnerabilities, including 24 linked to existing experimental or approved drugs and 4 previously uncharacterized targets in nucleotide and lipid synthesis. They further recapitulate a ceramide-linked metabolic stress signature shared across diverse chemotherapies. Mechanistic analysis traces the effects of BRCA1 loss to transcription-factor-mediated shifts in enzyme activity, outlining regulatory routes for network-level rewiring. Beyond ovarian cancer, this framework offers a generalizable blueprint for predicting metabolic vulnerabilities, drug responses, and adaptive mechanisms across diverse cancer and metabolic disease contexts. By coupling dynamic metabolism to therapeutic prediction, it delivers actionable hypotheses for biomarker discovery, patient stratification, target prioritization, and precision metabolic medicine.
Open AccessMulti-omics-driven kinetic modeling reveals metabolic adaptations and vulnerabilities in BRCA1-deficient ovarian cancer
BRCA1 -deficient ovarian cancer cells undergo extensive metabolic reprogramming, yet the network-level dynamics underlying their proliferation and treatment response remain poorly resolved. Here, we construct large-scale multi-omics-driven kinetic model populations of ovarian cancer metabolism to track how tumor cells adapt to changes in nutrient use, energy production, and metabolite dynamics over time. Across BRCA1 wild-type and mutant cells, these models expose distinct metabolic strategies shaped by transcriptional regulation and prioritize 28 enzyme-mediated vulnerabilities, including 24 linked to existing experimental or approved drugs and 4 previously uncharacterized targets in nucleotide and lipid synthesis. They further recapitulate a ceramide-linked metabolic stress signature shared across diverse chemotherapies. Mechanistic analysis traces the effects of BRCA1 loss to transcription-factor-mediated shifts in enzyme activity, outlining regulatory routes for network-level rewiring. Beyond ovarian cancer, this framework offers a generalizable blueprint for predicting metabolic vulnerabilities, drug responses, and adaptive mechanisms across diverse cancer and metabolic disease contexts. By coupling dynamic metabolism to therapeutic prediction, it delivers actionable hypotheses for biomarker discovery, patient stratification, target prioritization, and precision metabolic medicine.
- Open Access
- 18 June 2025
- Published by Elsevier in Metabolic Engineering
- Vol. 91, 430-441
Kinetic-model-guided engineering of multiple S. cerevisiae strains improves p-coumaric acid production
- 18 June 2025
- Published by Elsevier in Metabolic Engineering
- Vol. 91, 430-441
Open AccessKinetic-model-guided engineering of multiple S. cerevisiae strains improves p-coumaric acid production
- Open Access
- 22 April 2025
- Published by American Chemical Society (ACS) in ACS Synthetic Biology
- Vol. 14 (8) , 2906-2919
The Dawn of High-Throughput and Genome-Scale Kinetic Modeling: Recent Advances and Future Directions
Researchers have invested much effort into developing kinetic models due to their ability to capture dynamic behaviors, transient states, and regulatory mechanisms of metabolism, providing a detailed and realistic representation of cellular processes. Historically, the requirements for detailed parametrization and significant computational resources created barriers to their development and adoption for high-throughput studies. However, recent advancements, including the integration of machine learning with mechanistic metabolic models, the development of novel kinetic parameter databases, and the use of tailor-made parametrization strategies, are reshaping the field of kinetic modeling. In this Review, we discuss these developments and offer future directions, highlighting the potential of these advances to drive progress in systems and synthetic biology, metabolic engineering, and medical research at an unprecedented scale and pace.
- 22 April 2025
- Published by American Chemical Society (ACS) in ACS Synthetic Biology
- Vol. 14 (8) , 2906-2919
Open AccessThe Dawn of High-Throughput and Genome-Scale Kinetic Modeling: Recent Advances and Future Directions
Researchers have invested much effort into developing kinetic models due to their ability to capture dynamic behaviors, transient states, and regulatory mechanisms of metabolism, providing a detailed and realistic representation of cellular processes. Historically, the requirements for detailed parametrization and significant computational resources created barriers to their development and adoption for high-throughput studies. However, recent advancements, including the integration of machine learning with mechanistic metabolic models, the development of novel kinetic parameter databases, and the use of tailor-made parametrization strategies, are reshaping the field of kinetic modeling. In this Review, we discuss these developments and offer future directions, highlighting the potential of these advances to drive progress in systems and synthetic biology, metabolic engineering, and medical research at an unprecedented scale and pace.
- Open Access
Generative Approaches to Kinetic Parameter Inference in Metabolic Networks via Latent Space Exploration
Generative machine learning methods that utilize neural networks to parameterize large-scale and near-genome-scale kinetic models have yielded significant efficiency gains in model construction, paving the way for high-throughput dynamic metabolism studies in biomedical and biotechnological applications. Nevertheless, challenges remain in interpreting the outputs of generative neural networks and developing strategies to quickly adapt these networks to different organisms and physiological contexts without having to restart the modeling process from scratch. Here, we present a systematic framework for repurposing generative neural networks trained under one physiological context to build large-scale kinetic models tailored to another. We showcase the effectiveness of this framework through three case studies in Escherichia coli : (i) adjusting the speed of the dynamic response of aerobic metabolism, (ii) improving interpretability by identifying key enzymatic steps that limit the dynamic response speed of the metabolic models, and (iii) adapting a trained generator to capture the distinct dynamic behavior of anaerobic metabolism. To assess robustness and generalizability beyond E. coli , we extend our approach to large-scale kinetic models of Saccharomyces cerevisiae , systematically exploring latent-space–driven control of network dynamics across generators at different training stages and across multiple representative regions of the latent input space. Together, these results demonstrate that latent space exploration provides a transferable and computationally efficient strategy for controlling the dynamic behavior in large-scale kinetic models across species and physiological regimes. Given the growing adoption of generative neural networks in biological systems modeling, our approach has the potential to facilitate applications in personalized medicine and accelerate the high-throughput design of cell factories by streamlining model construction across diverse living organisms.
Open AccessGenerative Approaches to Kinetic Parameter Inference in Metabolic Networks via Latent Space Exploration
Generative machine learning methods that utilize neural networks to parameterize large-scale and near-genome-scale kinetic models have yielded significant efficiency gains in model construction, paving the way for high-throughput dynamic metabolism studies in biomedical and biotechnological applications. Nevertheless, challenges remain in interpreting the outputs of generative neural networks and developing strategies to quickly adapt these networks to different organisms and physiological contexts without having to restart the modeling process from scratch. Here, we present a systematic framework for repurposing generative neural networks trained under one physiological context to build large-scale kinetic models tailored to another. We showcase the effectiveness of this framework through three case studies in Escherichia coli : (i) adjusting the speed of the dynamic response of aerobic metabolism, (ii) improving interpretability by identifying key enzymatic steps that limit the dynamic response speed of the metabolic models, and (iii) adapting a trained generator to capture the distinct dynamic behavior of anaerobic metabolism. To assess robustness and generalizability beyond E. coli , we extend our approach to large-scale kinetic models of Saccharomyces cerevisiae , systematically exploring latent-space–driven control of network dynamics across generators at different training stages and across multiple representative regions of the latent input space. Together, these results demonstrate that latent space exploration provides a transferable and computationally efficient strategy for controlling the dynamic behavior in large-scale kinetic models across species and physiological regimes. Given the growing adoption of generative neural networks in biological systems modeling, our approach has the potential to facilitate applications in personalized medicine and accelerate the high-throughput design of cell factories by streamlining model construction across diverse living organisms.
- Open Access
Kinetic-model-guided engineering of multiple S. cerevisiae strains improves p -coumaric acid production
The use of kinetic models of metabolism in design-build-learn-test cycles is limited despite their potential to guide and accelerate the optimization of cell factories. This is primarily due to difficulties in constructing kinetic models capable of capturing the complexities of the fermentation conditions. Building on recent advances in kinetic-model-based strain design, we present the rational metabolic engineering of an S. cerevisiae strain designed to overproduce p -coumaric acid ( p -CA), an aromatic amino acid with valuable nutritional and therapeutic applications. To this end, we built nine kinetic models of an already engineered p -CA-producing strain by integrating different types of omics data and imposing physiological constraints pertinent to the strain. These nine models contained 297 mass balances involved in 303 reactions across four compartments and could reproduce the dynamic characteristics of the strain in batch fermentation simulations. We used constraint-based metabolic control analysis to generate combinatorial designs of 3 enzyme manipulations that could increase p-CA yield on glucose while ensuring that the resulting engineering strains did not deviate far from the reference phenotype. Among 39 unique designs, 10 proved robust across the phenotypic uncertainty of the models and could reliably increase p -CA yield in nonlinear simulations. We implemented these top 10 designs in a batch fermentation setting using a promoter-swapping strategy for down-regulations and plasmids for up-regulations. Eight out of the ten designs produced higher p -CA titers than the reference strain, with 19 – 32% increases at the end of fermentation. Importantly, these eight designs also maintained at least 90% of the growth of the reference strain; this indicates the importance of the phenotypic constraint. The high success rate of our in-silico designs in an experimental setting demonstrates the potential and utility of kinetic-model-based strain design. This work sets the foundation for accelerated design-build-test-learn cycles using large-scale kinetic models as a scaffold.
Open AccessKinetic-model-guided engineering of multiple S. cerevisiae strains improves p -coumaric acid production
The use of kinetic models of metabolism in design-build-learn-test cycles is limited despite their potential to guide and accelerate the optimization of cell factories. This is primarily due to difficulties in constructing kinetic models capable of capturing the complexities of the fermentation conditions. Building on recent advances in kinetic-model-based strain design, we present the rational metabolic engineering of an S. cerevisiae strain designed to overproduce p -coumaric acid ( p -CA), an aromatic amino acid with valuable nutritional and therapeutic applications. To this end, we built nine kinetic models of an already engineered p -CA-producing strain by integrating different types of omics data and imposing physiological constraints pertinent to the strain. These nine models contained 297 mass balances involved in 303 reactions across four compartments and could reproduce the dynamic characteristics of the strain in batch fermentation simulations. We used constraint-based metabolic control analysis to generate combinatorial designs of 3 enzyme manipulations that could increase p-CA yield on glucose while ensuring that the resulting engineering strains did not deviate far from the reference phenotype. Among 39 unique designs, 10 proved robust across the phenotypic uncertainty of the models and could reliably increase p -CA yield in nonlinear simulations. We implemented these top 10 designs in a batch fermentation setting using a promoter-swapping strategy for down-regulations and plasmids for up-regulations. Eight out of the ten designs produced higher p -CA titers than the reference strain, with 19 – 32% increases at the end of fermentation. Importantly, these eight designs also maintained at least 90% of the growth of the reference strain; this indicates the importance of the phenotypic constraint. The high success rate of our in-silico designs in an experimental setting demonstrates the potential and utility of kinetic-model-based strain design. This work sets the foundation for accelerated design-build-test-learn cycles using large-scale kinetic models as a scaffold.
- Open Access
- 30 August 2024
- Published by Springer Nature in Nature Catalysis
- Vol. 7 (10) , 1086-1098
Generative machine learning produces kinetic models that accurately characterize intracellular metabolic states
Generating large omics datasets has become routine for gaining insights into cellular processes, yet deciphering these datasets to determine metabolic states remains challenging. Kinetic models can help integrate omics data by explicitly linking metabolite concentrations, metabolic fluxes and enzyme levels. Nevertheless, determining the kinetic parameters that underlie cellular physiology poses notable obstacles to the widespread use of these mathematical representations of metabolism. Here we present RENAISSANCE, a generative machine learning framework for efficiently parameterizing large-scale kinetic models with dynamic properties matching experimental observations. Through seamless integration of diverse omics data and other relevant information, including extracellular medium composition, physicochemical data and expertise of domain specialists, RENAISSANCE accurately characterizes intracellular metabolic states in Escherichia coli. It also estimates missing kinetic parameters and reconciles them with sparse experimental data, substantially reducing parameter uncertainty and improving accuracy. This framework will be valuable for researchers studying metabolic variations involving changes in metabolite and enzyme levels and enzyme activity in health and biotechnology.
- 30 August 2024
- Published by Springer Nature in Nature Catalysis
- Vol. 7 (10) , 1086-1098
Open AccessGenerative machine learning produces kinetic models that accurately characterize intracellular metabolic states
Generating large omics datasets has become routine for gaining insights into cellular processes, yet deciphering these datasets to determine metabolic states remains challenging. Kinetic models can help integrate omics data by explicitly linking metabolite concentrations, metabolic fluxes and enzyme levels. Nevertheless, determining the kinetic parameters that underlie cellular physiology poses notable obstacles to the widespread use of these mathematical representations of metabolism. Here we present RENAISSANCE, a generative machine learning framework for efficiently parameterizing large-scale kinetic models with dynamic properties matching experimental observations. Through seamless integration of diverse omics data and other relevant information, including extracellular medium composition, physicochemical data and expertise of domain specialists, RENAISSANCE accurately characterizes intracellular metabolic states in Escherichia coli. It also estimates missing kinetic parameters and reconciles them with sparse experimental data, substantially reducing parameter uncertainty and improving accuracy. This framework will be valuable for researchers studying metabolic variations involving changes in metabolite and enzyme levels and enzyme activity in health and biotechnology.
- Open Access
- 24 January 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-14
Rational strain design with minimal phenotype perturbation
Devising genetic interventions for desired cellular phenotypes remains challenging regarding time and resources. Kinetic models can accelerate this task by simulating metabolic responses to genetic perturbations. However, exhaustive design evaluations with kinetic models are computationally impractical, especially when targeting multiple enzymes. Here, we introduce a framework for efficiently scouting the design space while respecting cellular physiological requirements. The framework employs mixed-integer linear programming and nonlinear simulations with large-scale nonlinear kinetic models to devise genetic interventions while accounting for the network effects of these perturbations. Importantly, it ensures the engineered strain’s robustness by maintaining its phenotype close to that of the reference strain. The framework, applied to improve the anthranilate production in E. coli, devises designs for experimental implementation, including eight previously experimentally validated targets. We expect this framework to play a crucial role in future design-build-test-learn cycles, significantly expediting the strain design compared to exhaustive design enumeration.
- 24 January 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-14
Open AccessRational strain design with minimal phenotype perturbation
Devising genetic interventions for desired cellular phenotypes remains challenging regarding time and resources. Kinetic models can accelerate this task by simulating metabolic responses to genetic perturbations. However, exhaustive design evaluations with kinetic models are computationally impractical, especially when targeting multiple enzymes. Here, we introduce a framework for efficiently scouting the design space while respecting cellular physiological requirements. The framework employs mixed-integer linear programming and nonlinear simulations with large-scale nonlinear kinetic models to devise genetic interventions while accounting for the network effects of these perturbations. Importantly, it ensures the engineered strain’s robustness by maintaining its phenotype close to that of the reference strain. The framework, applied to improve the anthranilate production in E. coli, devises designs for experimental implementation, including eight previously experimentally validated targets. We expect this framework to play a crucial role in future design-build-test-learn cycles, significantly expediting the strain design compared to exhaustive design enumeration.