YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
kcat-km final blend
(A) frozen ESM-2 35M (last + layer-8 mean pool, PCA-256) + Morgan counts / MACCS / RDKit descriptors -> LightGBM per target, 5 seed(s) averaged; plus Extra Trees on the same features. (B) ESM-2 35M, top 4 layers fine-tuned + molecule MLP + enzyme x substrate cross features, 4 seeds averaged (variant: rank). Trained on the organizer train set + leak-filtered public kinetics tables; no kinetics-pretrained models. Blend: w_nn (kcat, Km) = [0.5, 0.5]; Extra Trees weight 0.3. Grouped five-fold CV of this design: kcat 0.4751, Km 0.6872, mean 0.5811. One shared ESM pass for layers 0-7. CPU only. Run: python /app/predict.py --input INPUT --output OUTPUT