Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Objective-Guided Discrete Flow Matching for Controllable Biological Sequence Design

Tong Chen, Yinuo Zhang, Sophia Tang, Pranam Chatterjee

TL;DR

MOG-DFM introduces a general framework to steer pretrained discrete flow matching models toward Pareto-efficient generation across multiple scalar objectives in discrete biomolecule spaces. It combines rank-based local improvements with directional guidance using weight vectors drawn from the Das–Dennis lattice, plus an adaptive hypercone filtering mechanism and Euler sampling to produce sequences whose multi-objective scores lie near the Pareto front. The approach is instantiated with PepDFM for peptides and EnhancerDFM for enhancer DNA, and demonstrated on five-property peptide design and class/shape-guided DNA design, outperforming traditional multi-objective baselines and illustrating improved downstream docking, folding, and property scores. Together, MOG-DFM offers a scalable, controllable method for practical multi-property biomolecule design in discrete sequence spaces, with broad potential for therapeutics and synthetic biology.

Abstract

Designing biological sequences that satisfy multiple, often conflicting, functional and biophysical criteria remains a central challenge in biomolecule engineering. While discrete flow matching models have recently shown promise for efficient sampling in high-dimensional sequence spaces, existing approaches address only single objectives or require continuous embeddings that can distort discrete distributions. We present Multi-Objective-Guided Discrete Flow Matching (MOG-DFM), a general framework to steer any pretrained discrete flow matching generator toward Pareto-efficient trade-offs across multiple scalar objectives. At each sampling step, MOG-DFM computes a hybrid rank-directional score for candidate transitions and applies an adaptive hypercone filter to enforce consistent multi-objective progression. We also trained two unconditional discrete flow matching models, PepDFM for diverse peptide generation and EnhancerDFM for functional enhancer DNA generation, as base generation models for MOG-DFM. We demonstrate MOG-DFM's effectiveness in generating peptide binders optimized across five properties (hemolysis, non-fouling, solubility, half-life, and binding affinity), and in designing DNA sequences with specific enhancer classes and DNA shapes. In total, MOG-DFM proves to be a powerful tool for multi-property-guided biomolecule sequence design.

Multi-Objective-Guided Discrete Flow Matching for Controllable Biological Sequence Design

TL;DR

MOG-DFM introduces a general framework to steer pretrained discrete flow matching models toward Pareto-efficient generation across multiple scalar objectives in discrete biomolecule spaces. It combines rank-based local improvements with directional guidance using weight vectors drawn from the Das–Dennis lattice, plus an adaptive hypercone filtering mechanism and Euler sampling to produce sequences whose multi-objective scores lie near the Pareto front. The approach is instantiated with PepDFM for peptides and EnhancerDFM for enhancer DNA, and demonstrated on five-property peptide design and class/shape-guided DNA design, outperforming traditional multi-objective baselines and illustrating improved downstream docking, folding, and property scores. Together, MOG-DFM offers a scalable, controllable method for practical multi-property biomolecule design in discrete sequence spaces, with broad potential for therapeutics and synthetic biology.

Abstract

Designing biological sequences that satisfy multiple, often conflicting, functional and biophysical criteria remains a central challenge in biomolecule engineering. While discrete flow matching models have recently shown promise for efficient sampling in high-dimensional sequence spaces, existing approaches address only single objectives or require continuous embeddings that can distort discrete distributions. We present Multi-Objective-Guided Discrete Flow Matching (MOG-DFM), a general framework to steer any pretrained discrete flow matching generator toward Pareto-efficient trade-offs across multiple scalar objectives. At each sampling step, MOG-DFM computes a hybrid rank-directional score for candidate transitions and applies an adaptive hypercone filter to enforce consistent multi-objective progression. We also trained two unconditional discrete flow matching models, PepDFM for diverse peptide generation and EnhancerDFM for functional enhancer DNA generation, as base generation models for MOG-DFM. We demonstrate MOG-DFM's effectiveness in generating peptide binders optimized across five properties (hemolysis, non-fouling, solubility, half-life, and binding affinity), and in designing DNA sequences with specific enhancer classes and DNA shapes. In total, MOG-DFM proves to be a powerful tool for multi-property-guided biomolecule sequence design.
Paper Structure (27 sections, 30 equations, 5 figures, 11 tables, 1 algorithm)

This paper contains 27 sections, 30 equations, 5 figures, 11 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Discrete Flow Matching
  3. Multi-Objective Guided Discrete Flow Matching
  4. Step 0: Initialization and Weight Vector Generation
  5. Step 1: Guided Transition Scoring
  6. Step 2: Adaptive Hypercone Filtering
  7. Step 3: Euler Sampling
  8. Experiments
  9. PepDFM and EnhancerDFM Generate Diverse and Biologically Plausible Sequences
  10. MOG-DFM effectively balances each objective trade-off
  11. MOG-DFM generates peptide binders under five property guidance
  12. Hyperparameter Sensitivity Benchmark
  13. Adaptive Hypercone Filtering Enhances Multi-Objective Optimization
  14. MOG-DFM generates enhancer DNA of specific class with specified DNA shapes
  15. Conclusion
  16. ...and 12 more sections

Figures (5)

  • Figure 1: Visualization for MOG-DFM algorithm.
  • Figure 2: (A), (B) Complex structures of PDB 5AZ8 with a MOG-DFM-designed binder and its pre-existing binder. (C), (D) Complex structures of two target proteins without pre-existing binders (OX1R, EWS::FLI1) with MOG-DFM-designed binders. Five property scores are shown for each binder, along with the ipTM score from AlphaFold3 and docking score from AutoDock VINA. Interacting residues on the target are visualized. (E) Plots showing the mean scores for each property across the number of iterations during MOG-DFM's design of binders of length 12-aa for EWS::FLI1. (F) Density plots illustrating the distribution of predicted property scores for MOG-DFM-designed EWS::FLI1 binders of length 12 aa, compared to the peptides generated unconditionally by PepDFM. Please zoom in for better viewing.
  • Figure 3: (A) The Hamming distance of sampled peptides of different lengths to the peptides of the same length in the test set. (B) The Shannon Entropy of sampled peptides of different lengths to the peptides of the same length in the test set.
  • Figure 4: Complex structures of target proteins with pre-existing binders.(A)-(B) 1B8Q, (C)-(D) 1E6I, (E)-(F) 3IDJ, (G)-(H) 7JVS. Each panel shows the complex structure of the target with either a MOG-DFM-designed binder or its pre-existing binder. For each binder, five property scores are provided, as well as the ipTM score from AlphaFold3 and the docking score from AutoDock VINA. Interacting residues on the target are visualized.
  • Figure 5: Complex structures of target proteins without pre-existing binders.(A)-(C) AMHR2, (D)-(E) EWS::FLI1, (F) MYC, (G) DUSP12. Each panel shows the complex structure of the target with a MOG-DFM-designed binder. For each binder, five property scores are provided, as well as the ipTM score from AlphaFold3 and the docking score from AutoDock VINA. Interacting residues on the target are visualized.