MULTI-evolve (model-guided, universal, targeted installation of multi-mutants) is an end-to-end framework for engineering hyperactive multi-mutant proteins. MULTI-evolve trains neural networks to predict sequence fitness and uses them to propose combinatorial variants Then, MULTI-evolve generates MULTI-assembly mutagenic oligos for gene synthesis of the proposed top candidates. A protein language model zero-shot ensemble approach using ESM and ESM-IF enables nomination of promising single mutants for experimental testing. An interactive Streamlit web app and command-line interface make the full workflow accessible to the broader research community.
Github