Biomolecular sequence editing meets multi-objective optimization and feasibility
pCoMole: Pareto-Constrained Molecule Editing with Discrete Flows
Machine Learning
Summary
Designing new biomolecules like proteins or peptides often means changing their sequences to improve several features at once, while making sure they still work and can be made reliably. The authors created pCoMole, a method that guides step-by-step editing of molecule sequences to meet several goals and hard rules simultaneously. Their approach uses a mathematical trick to steer the editing process toward the best trade-offs and checks that the final sequences are valid. They tested it by making smaller versions of green fluorescent protein and Cas9 enzymes that still worked well, and shrinking drug-like peptides while keeping important properties.
What this means in practice
- •For protein engineers: Produce shortened protein variants that preserve essential functions and meet strict biochemical and manufacturability criteria.
- •For drug discovery teams: Optimize peptide-based drug candidates for multiple pharmacological properties while ensuring hard feasibility constraints are met.
Authors
Tong Chen, Maximilian Holsman, Lin Zhao, Pranam Chatterjee
Abstract
Biomolecular therapeutics often start from known sequences and require targeted editing to improve multiple properties while satisfying hard biochemical and manufacturability constraints. However, existing generative methods do not jointly support multi-objective optimization, hard feasibility, and sequence editing in discrete, variable-length biological spaces. In this work, we introduce Pareto-Constrained Molecule Editing (pCoMole), a framework built on discrete flow matching that steers a pre-trained Edit Flow toward user-specified preferences while enforcing terminal feasibility. pCoMole defines a feasibility-gated terminal distribution using an augmented Tchebycheff utility and realizes the resulting preference tilt through a Doob-h transform of the underlying edit process. To make this construction practical, we approximate the required harmonic function using short Monte Carlo rollouts over candidate edits, yielding an efficient guided editor with provable preference consistency. We validate pCoMole by shrinking GFP while retaining fluorescence-related properties, shortening diverse Cas9 orthologs while preserving PAM specificity, and compressing peptide binders into short peptidomimetics that optimize seven drug-related properties under hard constraints. In wet lab testing, two 229-residue pCoMole-designed eGFP variants retained clear green fluorescence in BL21 cells after 10 deletions, with either one or two substitutions. Together, pCoMole enables constraint-aware, Pareto-aligned editing of biomolecular sequences in discrete, variable-length spaces.