Deep learning to generate in silico chemical property libraries and candidate molecules for small molecule identification in complex samples

TL;DR

This paper introduces a deep learning framework using a variational autoencoder to generate in silico chemical property libraries and candidate molecules, significantly enhancing small molecule identification in complex samples.

Contribution

It develops a novel VAE-based method with multitask training to expand reference libraries and generate molecules with desired properties, overcoming current limitations.

Findings

Successfully predicts chemical properties from structure

Generates candidate molecules with specific properties

Enables rapid molecule identification and design

Abstract

Comprehensive and unambiguous identification of small molecules in complex samples will revolutionize our understanding of the role of metabolites in biological systems. Existing and emerging technologies have enabled measurement of chemical properties of molecules in complex mixtures and, in concert, are sensitive enough to resolve even stereoisomers. Despite these experimental advances, small molecule identification is inhibited by (i) chemical reference libraries representing <1% of known molecules, limiting the number of possible identifications, and (ii) the lack of a method to generate candidate matches directly from experimental features (i.e. without a library). To this end, we developed a variational autoencoder (VAE) to learn a continuous numerical, or latent, representation of molecular structure to expand reference libraries for small molecule identification. We extended the VAE to include a chemical property decoder, trained as a multitask network, in order to shape the latent representation such that it assembles according to desired chemical properties. The approach is unique in its application to small molecule identification, with its focus on m/z and CCS, paired with its training paradigm, which involved a cascade of transfer learning iterations. This allows the network to learn as much as possible at each stage, enabling success with progressively smaller datasets without overfitting. Once trained, the network can rapidly predict chemical properties directly from structure, as well as generate candidate structures with desired chemical properties. Additionally, the ability to generate novel molecules along manifolds, defined by chemical property analogues, positions DarkChem as highly useful in a number of application areas, including metabolomics and small molecule identification, drug discovery and design, chemical forensics, and beyond.