GraphBPE: Molecular Graphs Meet Byte-Pair Encoding

TL;DR

GraphBPE introduces a novel graph tokenization method inspired by Byte-Pair Encoding, improving molecular graph data preprocessing and enhancing model performance across various datasets and architectures.

Contribution

This paper presents GraphBPE, a new substructure tokenization technique for molecular graphs that is model-agnostic and improves data preprocessing in molecular machine learning.

Findings

GraphBPE boosts performance on small classification datasets.

It performs comparably with other tokenization methods across architectures.

Data preprocessing significantly impacts molecular graph model performance.

Abstract

With the increasing attention to molecular machine learning, various innovations have been made in designing better models or proposing more comprehensive benchmarks. However, less is studied on the data preprocessing schedule for molecular graphs, where a different view of the molecular graph could potentially boost the model's performance. Inspired by the Byte-Pair Encoding (BPE) algorithm, a subword tokenization method popularly adopted in Natural Language Processing, we propose GraphBPE, which tokenizes a molecular graph into different substructures and acts as a preprocessing schedule independent of the model architectures. Our experiments on 3 graph-level classification and 3 graph-level regression datasets show that data preprocessing could boost the performance of models for molecular graphs, and GraphBPE is effective for small classification datasets and it performs on par with other tokenization methods across different model architectures.