A computational framework named SKiP predicts catalytic turnover numbers and Michaelis constants directly from the three-dimensional structures of enzyme-substrate complexes, according to findings published in the Journal of Cheminformatics. Developed to overcome generalization and interpretability limitations in sequence-centric tools, the system provides a mechanistically driven approach to forecasting fundamental kinetic values in biocatalysis and enzyme engineering.
Predicting kinetic parameters has traditionally served as a primary bottleneck across industrial biotechnology and protein engineering workflows. Previous predictive methods relied predominantly on sequence-only pipelines, which frequently struggle to generalize across diverse proteins or provide clear mechanistic insights. To address these limitations, researchers Sowmya Ramaswamy Krishnan, Nishtha Pandey, and Arijit Roy structured SKiP to integrate explicit three-dimensional molecular configurations into kinetic modeling.
The architecture unites five distinct tiers of structural data into a unified representation based on a recently curated structural dataset. These components combine global enzyme descriptors, local interaction fingerprints, active site descriptors, substrate physicochemical descriptors, and reaction-specific features. By linking these complementary structural inputs, the system captures both overall macromolecular properties and localized catalytic interfaces critical to biological catalysis.
The model distinguishes itself from prior approaches through three core technical design elements. First, it explicitly incorporates the holo enzyme context, accounting for known cofactor interactions alongside pH-conditioned protonation states. These environmental determinants and active site conditions are typically treated only implicitly, if at all, within sequence-based pipelines, despite their significant influence on real catalytic rates.
Second, SKiP relies on explicitly defined physicochemical descriptors across the enzyme, substrate, and reaction components. This transparent parameterization facilitates rigorous analysis of the specific feature groups driving individual model estimations. Consequently, researchers can examine the chemical properties contributing to each kinetic prediction rather than relying on uninterpretable statistical associations.
Third, the framework improves predictive generalizability over previously unseen data by incorporating dual uncertainty quantification. SKiP assesses both aleatoric and epistemic uncertainty, allowing the platform to distinguish high-confidence estimates from low-confidence projections at the level of individual data points. This uncertainty awareness provides an explicit reliability metric before computational predictions advance into laboratory testing.
Targeted validation studies confirmed that SKiP reproduces experimentally observed kinetic behaviors across distinct enzyme families. The framework accurately captured the pH-dependent catalytic response in a hyperthermophilic beta-glucosidase. Furthermore, it discriminated between enantiomeric substrates within stereospecific lactate dehydrogenases, reproducing known active site interaction patterns and confirming that structural inputs improve generalizability across diverse enzymes, mutants, and substrates.
Collectively, the authors report that embedding explicit structural information, interpretable interaction fingerprints, and calibrated uncertainty estimates creates a practical foundation for biocatalyst discovery. By offering reliable parameter estimations for complex biological systems, the framework assists researchers in prioritizing candidate enzymes and guiding rational engineering campaigns across biocatalysis workflows.
Subject of Research: Structure-based prediction of enzyme kinetic parameters
Article Title: SKiP: a Structure-oriented Kinetics Prediction suite for enzyme-substrate complexes
Article References: Krishnan, S. R., Pandey, N., & Roy, A. (2026). SKiP: a Structure-oriented Kinetics Prediction suite for enzyme-substrate complexes. Journal of Cheminformatics. https://doi.org/10.1186/s13321-026-01333-z
Image Credits: AI Generated
DOI: 10.1186/s13321-026-01333-z
Keywords: enzyme kinetics, turnover number, Michaelis constant, enzyme-substrate complexes, biocatalysis, enzyme engineering, structural bioinformatics, computational chemistry, cheminformatics, holo enzyme context, pH-dependent catalysis, active site descriptors, aleatoric uncertainty, epistemic uncertainty, beta-glucosidase, lactate dehydrogenase, stereospecificity, molecular interaction fingerprints
Cite Scienmag News
Priscilla Baird. (November 17, 2026). Structure-Aware Model Predicts Enzyme Kinetic Parameters Directly From Complex Structures. Scienmag. https://scienmag.com/structure-aware-model-predicts-enzyme-kinetic-parameters-directly-from-complex-structures/
Priscilla Baird. "Structure-Aware Model Predicts Enzyme Kinetic Parameters Directly From Complex Structures." Scienmag, 17 November 2026, https://scienmag.com/structure-aware-model-predicts-enzyme-kinetic-parameters-directly-from-complex-structures/. Accessed 17 November 2026.
Priscilla Baird. "Structure-Aware Model Predicts Enzyme Kinetic Parameters Directly From Complex Structures." Scienmag. November 17, 2026. https://scienmag.com/structure-aware-model-predicts-enzyme-kinetic-parameters-directly-from-complex-structures/








