A machine learning approach to predict metabolic pathway dynamics from time-series multiomics data
Problem
Bioengineers could not accurately predict how modifying an organism's genotype changes metabolic pathway behavior, and classical kinetic models are slow and expertise-intensive to build.
Why it matters
Unpredictable biological behavior is a fundamental hurdle to engineering useful biological systems.
What the AI did
A machine learning method using multiomics time-series data (proteomics, metabolomics) predicted pathway dynamics in an automated fashion, improving with more data.
Results
Outperforms a classical kinetic model; qualitative and quantitative predictions produced (no numeric metrics stated)
Impact
Peer-reviewed demonstration that ML can automate pathway-dynamics prediction and guide bioengineering.
Record
| Venue | npj Systems Biology and Applications |
|---|---|
| Published | 2018 |
| Work type | article |
| Authors | Zak Costello, Héctor García Martín |
| Citations | 242 (OpenAlex, at import time) |
| References | 94 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41540-018-0054-3 |
| Open access | gold |
| Keywords | machine learning, multiomics, metabolic pathways, synthetic biology, kinetic models |
Importance rating
significant — thousands to millions affected
Abstract calls inability to predict biological behavior after genotype modification a fundamental hurdle to engineering biological systems; kinetic models slow and expertise-heavy. No scale stated.