Computational variant prioritization pipeline for herbicide tolerance
A computational pipeline that generates ranked shortlists of protein variants for herbicide tolerance by integrating ligand docking, protein language model fitness predictions, stability scoring, and affinity co-folding. Reduces experimental testing by 70–82% across benchmarked target systems.
Overview
This computational variant-prioritization pipeline accelerates herbicide tolerance engineering by converting a target protein and its herbicide into a ranked shortlist of testable mutants. It addresses the critical bottleneck between computational design and experimental validation: instead of testing hundreds of variants, researchers receive a focused list of high-priority candidates with per-candidate evidence. It supports herbicide resistance research and protein engineering with reduced experimental effort.
Technical specifications
Pipeline workflow:
- Enumerates all single mutants within the binding pocket and rim (≤12 Å of the ligand), typically 100–1,500 candidates per target
- Scores candidates via four complementary layers: ligand docking, ESM-family protein language model zero-shot fitness, ThermoMPNN stability (ΔΔG), and Boltz-2 co-folding/affinity scoring with the herbicide
- Fuses per-layer shortlists and applies an evolutionary prior module mined from 27 known resistance mutations, weighted by ligand distance and charge conservation
- Integrates active learning via the SynbioAI platform's MultiEvolve and Active-Learning-assisted Directed Evolution (ALDE) models
Benchmark results (26 known mutations):
- AHAS/imazaquin: 7/9 recovered from 378 candidates
- psbA/terbutryn: 6/8 recovered from 494 candidates
- EPSPS/glyphosate: 3/4 recovered from 645 candidates
- ACCase/haloxyfop: 1/5 recovered from 890 candidates
Shortlists reduce lab testing by 70–82% per target.
Technology readiness level
The pipeline is at a validated prototype stage, benchmarked against four target/herbicide systems with known resistance mutations. Informed prioritization recovers a substantial fraction of functional variants while dramatically reducing the experimental search space.
A pilot validation phase is available for new target/herbicide systems, delivering a ranked shortlist with evidence annotations within four weeks. Optional extensions include molecular dynamics simulation and active learning round setup.
About Atria University
Atria University is a young, private, UGC-recognized STEM university in Bengaluru with small cohorts and an applied, industry-oriented character. Its curriculum is co-created with corporate partners through engineering sprints, practitioner-led learning, internships, and access to industrial facilities. The research ecosystem combines Centres of Excellence, advanced laboratories, real-world testbeds, and Atria Beyonder Labs, a Labs-as-a-Service platform supporting prototyping, validation, deployment, and commercialization. Located in Bengaluru’s technology cluster and backed by the Atria Group, the university offers connections to companies, infrastructure, and entrepreneurial networks. Research funding and collaborations are linked to national organizations including the Ministry of Earth Sciences, ISRO, and the Anusandhan National Research Foundation.
delivered to your inbox.