Pre-trained Molecular Language Models with Random Functional Group Masking

TL;DR

This paper introduces \\ours{}, a SMILES-based molecular language model that uses random functional group masking during pre-training to improve structure-aware molecular property prediction, outperforming existing models on multiple benchmarks.

Contribution

The paper proposes a novel SMILES-based pre-training method with functional group masking to enhance structure learning in molecular language models.

Findings

 \\ours{} outperforms existing models on 9 of 11 benchmark tasks.

 It demonstrates robustness and superior performance across diverse chemical property predictions.

 Functional group masking improves the model's ability to infer molecular structures.

Abstract

Recent advancements in computational chemistry have leveraged the power of trans-former-based language models, such as MoLFormer, pre-trained using a vast amount of simplified molecular-input line-entry system (SMILES) sequences, to understand and predict molecular properties and activities, a critical step in fields like drug discovery and materials science. To further improve performance, researchers have introduced graph neural networks with graph-based molecular representations, such as GEM, incorporating the topology, geometry, 2D or even 3D structures of molecules into pre-training. While most of molecular graphs in existing studies were automatically converted from SMILES sequences, it is to assume that transformer-based language models might be able to implicitly learn structure-aware representations from SMILES sequences. In this paper, we propose \ours{} -- a SMILES-based \underline{\em M}olecular \underline{\em L}anguage \underline{\em M}odel, which randomly masking SMILES subsequences corresponding to specific molecular \underline{\em F}unctional \underline{\em G}roups to incorporate structure information of atoms during the pre-training phase. This technique aims to compel the model to better infer molecular structures and properties, thus enhancing its predictive capabilities. Extensive experimental evaluations across 11 benchmark classification and regression tasks in the chemical domain demonstrate the robustness and superiority of \ours{}. Our findings reveal that \ours{} outperforms existing pre-training models, either based on SMILES or graphs, in 9 out of the 11 downstream tasks, ranking as a close second in the remaining ones.