Combining mechanistic and machine learning models for predictive engineering and optimization of tryptophan metabolism
Problem
Predicting which genetic designs improve yeast tryptophan production is difficult, making metabolic engineering slow and inefficient.
Why it matters
Engineering aromatic amino acid metabolism in living systems is a central goal of industrial metabolic engineering.
What the AI did
A genome-scale mechanistic model pinpointed engineering targets, and machine learning algorithms trained on biosensor-screened library data recommended pathway designs.
Results
Best ML-guided designs improved tryptophan titer by up to 74% and productivity by up to 43% versus best training designs, from a single data-generation cycle.
Impact
Shows one ML-guided design cycle can yield large percentage gains in yeast tryptophan production.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Jie Zhang, Søren D. Petersen, Tijana Radivojević, Andrés Ramirez, Andrés Pérez-Manríquez, Eduardo Abeliuk, Benjamín J. Sánchez, Zak Costello et al. |
| Citations | 282 (OpenAlex, at import time) |
| References | 82 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-020-17910-1 |
| Open access | gold |
| Keywords | machine learning, metabolic engineering, tryptophan, yeast, genome-scale model |
Importance rating
real but narrower advance
Abstract shows large tryptophan gains in yeast and frames the method as effectively directing metabolic engineering, but states no scale or end-use impact; industrial biotech importance is inferred.