Simulated Design–Build–Test–Learn Cycles for Consistent Comparison of Machine Learning Methods in Metabolic Engineering
Problem
No framework existed to consistently test machine learning methods across multiple design-build-test-learn cycles in combinatorial strain optimization.
Why it matters
Combinatorial pathway optimization suffers explosive search spaces, making efficient strain development costly and slow.
What the AI did
Authors built a mechanistic kinetic model-based simulation framework to benchmark ML methods proposing designs across iterative DBTL cycles.
Results
Gradient boosting and random forest outperform other models; baseline comparison truncated in abstract.
Impact
Provides an in-silico benchmarking framework; no wet-lab validation stated.
Record
| Venue | ACS Synthetic Biology |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Paul van Lent, Joep P. J. Schmitz, Thomas Abeel |
| Citations | 54 (OpenAlex, at import time) |
| References | 59 |
| Model or method | gradient boosting, random forest |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1021/acssynbio.3c00186 |
| Open access | hybrid |
| Keywords | DBTL cycles, machine learning, metabolic engineering, gradient boosting, random forest, benchmarking |
Importance rating
real but narrower advance
Abstract calls combinatorial pathway optimization 'an important tool' and notes combinatorial explosions, but the work is an in-silico benchmarking framework with no stated real-world scale or wet-lab validation.