Data for "GENKI: A generative framework for scalable and robust metabolic kinetic modeling"

This dataset accompanies the paper "GENKI: A generative framework for scalable and robust metabolic kinetic modeling" (Xenios et al., 2025). GENKI uses a conditional Variational Autoencoder (cVAE) trained on ORACLE-sampled kinetic parameter ensembles to generate large, kinetically feasible Km parameter sets for metabolic models. The generated ensembles are screened for dynamic stability and evaluated against experimental flux data using a zeta-score KPI. This deposit contains: ORACLE Km ensembles used as training data for E. coli and S. cerevisiae models Per-perturbation ORACLE flux reference CSVs (E. coli: PGI, PGL, RPI, TALA, PYK, G6PDH2r; yeast: nanoaerobic → microaerobic transition) Pre-generated Km parameter sets from the trained cVAE (one per figure scenario) for direct figure reproduction without retraining Code to reproduce all paper figures is available at: https://github.com/stefanosxenios/GENKI