SOAPI: Siamese-guided generation of Off-Target-Avoiding Protein Interactions

Sophia Vincoff, Oscar Davis, Alexander Tong, Joey Bose, Pranam Chatterjee

Track: Full Paper Track

Keywords: Siamese neural network, discrete diffusion sampling, protein design

TL;DR: SOAPI integrates a Siamese protein language model with masked diffusion to generate highly specific protein binders while minimizing off-target interactions.

Abstract: Therapeutics that modulate pathogenic proteins while avoiding off-target interactions are essential for effective drug design. However, designing binders that selectively engage a target protein while minimizing interactions with structurally or functionally similar proteins remains a major challenge. To address this, we introduce Siamese-guided strategy for the generation of Off target-Avoiding Protein Interactions, termed SOAPI. SOAPI leverages a Siamese protein language model with an adaptive Log-Sum-Exp Decoy Loss to enforce specificity by embedding fusion-specific binders close to their target while maintaining separation from off-targets. These optimized embeddings then guide a diffusion protein language model (DPLM), which generates binders using soft-value-based decoding (SVDD) and Sequential Monte Carlo resampling to iteratively refine candidates. In silico validation demonstrates strong target affinity and significant off-target avoidance, highlighting SOAPI’s potential for generating precise and selective protein interactions.

Attendance: Sophia Vincoff

Submission Number: 54