Transfer Learning for Molecular Property Predictions from Small Data Sets

TL;DR

This paper explores transfer learning to enhance molecular property prediction on small datasets, using large pre-training datasets and fine-tuning, with notable success on certain properties like HOMO-LUMO gaps.

Contribution

It introduces a transfer learning strategy for small data molecular property prediction, leveraging large datasets for pre-training to improve accuracy after fine-tuning.

Findings

Pre-training improves prediction accuracy for HOMO-LUMO gaps.

Transfer learning less effective for complex properties like solvation energies.

Fewer pre-training data points can sometimes yield better fine-tuned models.

Abstract

Machine learning has emerged as a new tool in chemistry to bypass expensive experiments or quantum-chemical calculations, for example, in high-throughput screening applications. However, many machine learning studies rely on small data sets, making it difficult to efficiently implement powerful deep learning architectures such as message passing neural networks. In this study, we benchmark common machine learning models for the prediction of molecular properties on two small data sets, for which the best results are obtained with the message passing neural network PaiNN, as well as SOAP molecular descriptors concatenated to a set of simple molecular descriptors tailored to gradient boosting with regression trees. To further improve the predictive capabilities of PaiNN, we present a transfer learning strategy that uses large data sets to pre-train the respective models and allows to obtain more accurate models after fine-tuning on the original data sets. The pre-training labels are obtained from computationally cheap ab initio or semi-empirical models and both data sets are normalized to mean zero and standard deviation one to align the labels' distributions. This study covers two small chemistry data sets, the Harvard Organic Photovoltaics data set (HOPV, HOMO-LUMO-gaps), for which excellent results are obtained, and on the Freesolv data set (solvation energies), where this method is less successful, probably due to a complex underlying learning task and the dissimilar methods used to obtain pre-training and fine-tuning labels. Finally, we find that for the HOPV data set, the final training results do not improve monotonically with the size of the pre-training data set, but pre-training with fewer data points can lead to more biased pre-trained models and higher accuracy after fine-tuning.