Dr.VAE: Drug Response Variational Autoencoder

TL;DR

This paper introduces Dr.VAE, a deep generative model based on Variational Autoencoders, to enhance drug response prediction accuracy by modeling gene state changes due to drug treatment.

Contribution

The paper presents a novel semi-supervised VAE model, Dr.VAE, that captures gene state changes and improves drug response prediction over existing benchmarks.

Findings

Outperforms benchmarks with 3-11% higher AUROC

Achieves 2-30% better AUPR

Joint training improves model performance

Abstract

We present two deep generative models based on Variational Autoencoders to improve the accuracy of drug response prediction. Our models, Perturbation Variational Autoencoder and its semi-supervised extension, Drug Response Variational Autoencoder (Dr.VAE), learn latent representation of the underlying gene states before and after drug application that depend on: (i) drug-induced biological change of each gene and (ii) overall treatment response outcome. Our VAE-based models outperform the current published benchmarks in the field by anywhere from 3 to 11% AUROC and 2 to 30% AUPR. In addition, we found that better reconstruction accuracy does not necessarily lead to improvement in classification accuracy and that jointly trained models perform better than models that minimize reconstruction error independently.