Deep learning-based kcat prediction enables improved enzyme-constrained model reconstruction
Problem
Experimentally measured enzyme turnover numbers (kcat) are sparse and noisy, limiting genome-scale metabolic modeling across organisms.
Why it matters
kcat values are key to understanding cellular metabolism, proteome allocation and physiological diversity.
What the AI did
DLKcat, a deep learning approach, predicts kcat from substrate structures and protein sequences; a Bayesian pipeline then parameterizes enzyme-constrained genome-scale metabolic models from the predictions.
Results
Predicted genome-scale kcat for 300+ yeast species; resulting models outperformed previous enzyme-constrained models in predicting phenotypes and proteomes; captures kcat changes in mutated enzymes.
Impact
Enabled high-throughput kcat prediction and improved phenotype/proteome prediction across hundreds of yeast species.
Record
| Venue | Nature Catalysis |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Feiran Li, Le Yuan, Hongzhong Lu, Gang Li, Yu Chen, Martin K. M. Engqvist, Eduard Johannes Kerkhoven, JENS KVIST NIELSEN |
| Citations | 460 (OpenAlex, at import time) |
| References | 78 |
| Model or method | DLKcat (deep learning) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41929-022-00798-z |
| Open access | hybrid |
| Keywords | kcat prediction, deep learning, enzyme kinetics, metabolic models, Bayesian pipeline, yeast |
Importance rating
significant — thousands to millions affected
kcat is key to understanding cellular metabolism, proteome allocation and physiological diversity; enables high-throughput prediction for any organism, applied to 300+ yeast species.