Predicting drug polypharmacology from cell morphology readouts using variational autoencoder latent space arithmetic

A variational autoencoder (VAE) is a machine learning algorithm, useful for generating a compressed and interpretable latent space. These representations have been generated from various biomedical data types and can be used to produce realistic-looking simulated data. However, standard vanilla VAEs suffer from entangled and uninformative latent spaces, which can be mitigated using other types of VAEs such as β-VAE and MMD-VAE. In this project, we evaluated the ability of VAEs to learn cell morphology characteristics derived from cell images. We trained and evaluated these three VAE variants—Vanilla VAE, β-VAE, and MMD-VAE—on cell morphology readouts and explored the generative capacity of each model to predict compound polypharmacology (the interactions of a drug with more than one target) using an approach called latent space arithmetic (LSA). To test the generalizability of the strategy, we also trained these VAEs using gene expression data of the same compound perturbations and found that gene expression provides complementary information. We found that the β-VAE and MMD-VAE disentangle morphology signals and reveal a more interpretable latent space. We reliably simulated morphology and gene expression readouts from certain compounds thereby predicting cell states perturbed with compounds of known polypharmacology. Inferring cell state for specific drug mechanisms could aid researchers in developing and identifying targeted therapeutics and categorizing off-target effects in the future.Author summary: We train machine learning algorithms to identify patterns of drug activity from cell morphology readouts. Known as variational autoencoders (VAE), these algorithms are unsupervised, meaning that they do not require any additional information other than the input data to learn. We train and systematically evaluate three different kinds of VAEs, each that learn different patterns, and we document performance and interpretability tradeoffs. In a comprehensive evaluation, we learn that an approach called latent space arithmetic (LSA) can predict cell states of compounds that interact with multiple targets and mechanisms, a well-known phenomenon known as drug polypharmacology. Importantly, we train other VAEs using a gene expression assay known as L1000 and compare performance to cell morphology VAEs. Gene expression and cell morphology are the two most common types of data for modeling cells. We discover that modeling cell morphology requires vastly different VAE parameters and architectures, and, importantly, that the data types are complementary; they predict polypharmacology of different compounds. Our benchmark and publicly available software will enable future VAE modeling improvements, and our drug polypharmacology predictions demonstrate that we can model potential off-target effects in drugs, which is an important step in drug discovery pipelines.