AffinityFlow: Guided Flows for Antibody Affinity Maturation

TL;DR

AffinityFlow tackles sequence-only antibody affinity maturation by marrying AlphaFlow-based structure generation with predictor-guided sampling in an alternating optimization loop. A co-teaching module leverages noisy biophysical energies to refine both sequence- and structure-based affinity predictors, enabling data-efficient learning. The framework enables targeted CDR mutations via inverse folding and restricted predictor guidance, achieving state-of-the-art improvements in functionality, specificity, and rationality on sdAb maturation benchmarks. This approach promises practical impact for rapid, antigen-informed antibody design while highlighting the need for careful governance of powerful design methods.

Abstract

Antibodies are widely used as therapeutics, but their development requires costly affinity maturation, involving iterative mutations to enhance binding affinity.This paper explores a sequence-only scenario for affinity maturation, using solely antibody and antigen sequences. Recently AlphaFlow wraps AlphaFold within flow matching to generate diverse protein structures, enabling a sequence-conditioned generative model of structure. Building on this, we propose an alternating optimization framework that (1) fixes the sequence to guide structure generation toward high binding affinity using a structure-based affinity predictor, then (2) applies inverse folding to create sequence mutations, refined by a sequence-based affinity predictor for post selection. A key challenge is the lack of labeled data for training both predictors. To address this, we develop a co-teaching module that incorporates valuable information from noisy biophysical energies into predictor refinement. The sequence-based predictor selects consensus samples to teach the structure-based predictor, and vice versa. Our method, AffinityFlow, achieves state-of-the-art performance in affinity maturation experiments. We plan to open-source our code after acceptance.

Figures (5)

• Figure 1: Illustration of alternating optimization.
• Figure 2: Illustration of co-teaching.
• Figure 3: Visualizations of model-generated antibody structures bound to the SARS-CoV-2 RBD (a) Relative to a fixed antigen, most model-generated antibodies (green) are predicted to bind with a noticeable rotation in binding pose compared to the WT conformation (blue). (b) Our model suggests several mutations frequently, in particular Ala105Pro may stabilize the CDR loop. (c) The buried Lys99Trp mutation interacts with multiple other aromatic residues across the interface.
• Figure 4: The antibody metrics versus scaling factor $\gamma$, normalized to their values at $\gamma$to those with $\gamma=5.0$.
• Figure 5: The three antibody metrics versus scaling factor $T$, normalized to their values at $T=3$.