Generalizable, Fast, and Accurate DeepQSPR with fastprop

TL;DR

This paper introduces fastprop, a DeepQSPR framework that uses molecular descriptors to achieve faster, more accurate property predictions across diverse datasets, surpassing learned representations.

Contribution

fastprop is a novel DeepQSPR method that combines molecular descriptors with efficient algorithms for superior speed and accuracy in property prediction.

Findings

fastprop outperforms learned representations on multiple datasets

fastprop achieves faster training and inference times

fastprop maintains high accuracy across diverse molecular datasets

Abstract

Quantitative Structure Property Relationship studies aim to define a mapping between molecular structure and arbitrary quantities of interest. This was historically accomplished via the development of descriptors which requires significant domain expertise and struggles to generalize. Thus the field has morphed into Molecular Property Prediction and been given over to learned representations which are highly generalizable. The paper introduces fastprop, a DeepQSPR framework which uses a cogent set of molecular level descriptors to meet and exceed the performance of learned representations on diverse datasets in dramatically less time. fastprop is freely available on github at github.com/JacksonBurns/fastprop.