Using Molecular Embeddings in QSAR Modeling: Does it Make a Difference?

TL;DR

This study systematically compares various molecular embedding techniques with traditional representations in QSAR modeling, revealing that embeddings do not significantly outperform traditional methods in predictive tasks.

Contribution

The paper provides a comprehensive experimental comparison of five molecular embedding methods against traditional descriptors and fingerprints in QSAR scenarios.

Findings

Molecular embeddings do not significantly outperform traditional representations in QSAR tasks.

Supervised embeddings are competitive with traditional methods, while unsupervised embeddings tend to perform worse.

A large-scale evaluation with over 25,000 models highlights the need for careful selection of molecular representations.

Abstract

With the consolidation of deep learning in drug discovery, several novel algorithms for learning molecular representations have been proposed. Despite the interest of the community in developing new methods for learning molecular embeddings and their theoretical benefits, comparing molecular embeddings with each other and with traditional representations is not straightforward, which in turn hinders the process of choosing a suitable representation for QSAR modeling. A reason behind this issue is the difficulty of conducting a fair and thorough comparison of the different existing embedding approaches, which requires numerous experiments on various datasets and training scenarios. To close this gap, we reviewed the literature on methods for molecular embeddings and reproduced three unsupervised and two supervised molecular embedding techniques recently proposed in the literature. We compared these five methods concerning their performance in QSAR scenarios using different classification and regression datasets. We also compared these representations to traditional molecular representations, namely molecular descriptors and fingerprints. As opposed to the expected outcome, our experimental setup consisting of over 25,000 trained models and statistical tests revealed that the predictive performance using molecular embeddings did not significantly surpass that of traditional representations. While supervised embeddings yielded competitive results compared to those using traditional molecular representations, unsupervised embeddings tended to perform worse than traditional representations. Our results highlight the need for conducting a careful comparison and analysis of the different embedding techniques prior to using them in drug design tasks, and motivate a discussion about the potential of molecular embeddings in computer-aided drug design.