Ljubiša Mišković

The paper introduces METEOR-K, a framework that incorporates metabolite turnover rates as constraints to improve the physiological consistency of steady-state metabolic models.

This paper presents a framework for adapting generative neural networks to build kinetic models for different organisms and physiological contexts by exploring latent space.

This study builds kinetic models of ovarian cancer metabolism to identify metabolic vulnerabilities and drug targets in BRCA1-deficient cells.

The paper uses kinetic models to guide metabolic engineering of S. cerevisiae strains, resulting in improved p-coumaric acid production with minimal impact on growth rate.

This paper introduces RENAISSANCE, a generative machine learning framework that accurately characterizes intracellular metabolic states by parameterizing kinetic models using omics data and experimental observations.

This paper introduces a framework for rational strain design in E. coli that minimizes phenotype perturbation using mixed-integer linear programming and kinetic models to improve anthranilate production.

BridgIT+ predicts protein-reaction associations by leveraging enzyme promiscuity and sequence profiles, outperforming existing methods on E. coli and predicted reactions.

This paper introduces REKINDLE, a deep learning framework that generates kinetic models of metabolic networks with dynamic properties, demonstrating its efficiency and versatility through three studies.

This paper develops yETFL, an extended ME-model for Saccharomyces cerevisiae that integrates expression constraints and reaction thermodynamics, demonstrating its predictive ability for growth rate, essential genes, and overflow metabolism.

This paper introduces a constraint-based framework, Network Response Analysis (NRA), to improve metabolic engineering by incorporating physiological constraints and genome editing limitations.

The paper develops large-scale kinetic models of Pseudomonas putida to predict metabolic responses and design metabolic engineering strategies for improved biofuel and biochemical production.

This paper introduces iSCHRUNK, a method combining Monte Carlo sampling and machine learning to reduce uncertainty in kinetic models, identifying key parameters for desired metabolic behavior.

The paper develops a method to model uncertainty in enzyme kinetics by expressing kinetic responses through thermodynamic equilibrium displacements and enzyme concentrations, enabling more efficient parameter sampling for metabolic models.

BridgIT assigns enzyme sequences to orphan and novel reactions by comparing reactive site structures, predicting sequences for 93% of reactions with three connecting bonds and nearly all with six.

The paper analyzes how uncertainty in intracellular fluxes and concentrations affects kinetic models of E. coli metabolism and proposes a workflow to improve prediction reliability.

This paper uses retrobiosynthesis to discover and evaluate novel biosynthetic pathways for producing five MEK precursors in E. coli.