Boltz is a Strong Baseline for Atom-level Representation Learning

TL;DR

This paper evaluates Boltz, a protein-focused model operating at atom-level granularity, for small-molecule tasks, demonstrating its competitive performance and establishing it as a strong baseline for atom-level molecular representation learning.

Contribution

The study explores Boltz's atom-level representations for small-molecule tasks, revealing its effectiveness and positioning it as a new strong baseline in the field.

Findings

Boltz performs competitively on ADMET property prediction.

Boltz is effective for molecular generation and optimization.

Protein-centric models can capture transferable chemical physics.

Abstract

Foundation models in molecular learning have advanced along two parallel tracks: protein models, which typically utilize evolutionary information to learn amino acid-level representations for folding, and small-molecule models, which focus on learning atom-level representations for property prediction tasks such as ADMET. Notably, cutting-edge protein-centric models such as Boltz now operate at atom-level granularity for protein-ligand co-folding, yet their atom-level expressiveness for small-molecule tasks remains unexplored. A key open question is whether these protein co-folding models capture transferable chemical physics or rely on protein evolutionary signals, which would limit their utility for small-molecule tasks. In this work, we investigate the quality of Boltz atom-level representations across diverse small-molecule benchmarks. Our results show that Boltz is competitive with specialized baselines on ADMET property prediction tasks and effective for molecular generation and optimization. These findings suggest that the representational capacity of cutting-edge protein-centric models has been underexplored and position Boltz as a strong baseline for atom-level representation learning for small molecules.