High-throughput computational design of herbicide-tolerant protein variants

Overview

This computational workflow rapidly prioritizes herbicide-tolerant protein variants that preserve overall protein structure while reducing herbicide binding and supporting expression. It narrows the enormous mutational search space before experimental build and screening.

By combining machine learning-based sequence design, statistical analysis of mutational tolerance, and molecular docking, the workflow produces a focused, experimentally tractable variant set with a computational rationale for both structure preservation and altered ligand interaction. Applications include engineering herbicide tolerance into proteins for crop protection and production systems.

Technical specifications

The workflow integrates several computational methods:

• LigandMPNN-based substitution scoring generates residue-level amino-acid substitution probabilities, with and without ligand awareness, to identify structurally permissive, ligand-sensitive positions.
• Shannon entropy analysis flags positions whose mutational tolerance changes significantly in the presence of the herbicide.
• Molecular docking evaluates candidate single and combinatorial mutations, ranking them by predicted binding affinity relative to the reference protein.
• Expression-relevant scoring extends the ranking to include properties important for the target production system.

The output is a prioritized, experimentally tractable variant set rather than an impractically large sequence library.

Technology readiness level

The underlying approach has been applied in the context of drug resistance, providing an initial basis for feasibility (unpublished). The planned development path includes benchmarking predictions against available experimental binding data, testing selected unexplored variants, and refining the scoring strategy through an iterative design-build-test cycle. The workflow is at an early-to-mid stage of development: core methods are established, with target-specific benchmarking and experimental validation planned as next steps.