Predicting chemotherapy response using a variational autoencoder approach

Abstract

Background Multiple studies have shown the utility of transcriptome-wide RNA-seq profiles as features for machine learning-based prediction of response to chemotherapy in cancer. While tumor transcriptome profiles are publicly available for thousands of tumors for many cancer types, a relatively modest number of tumor profiles are clinically annotated for response to chemotherapy. The paucity of labeled examples and the high dimension of the feature data limit performance for predicting therapeutic response using fully-supervised classification methods. Recently, multiple studies have established the utility of a deep neural network approach, the variational autoencoder (VAE), for generating meaningful latent features from original data. Here, we report the first study of a semi-supervised approach using VAE-encoded tumor transcriptome features and regularized gradient boosted decision trees (XGBoost) to predict chemotherapy drug response for five cancer types: colon, pancreatic, bladder, breast, and sarcoma. Results We found: (1) VAE-encoding of the tumor transcriptome preserves the cancer type identity of the tumor, suggesting preservation of biologically relevant information; and (2) as a feature-set for supervised classification to predict response-to-chemotherapy, the unsupervised VAE encoding of the tumor’s gene expression profile leads to better area under the receiver operating characteristic curve and area under the precision-recall curve classification performance than the original gene expression profile or the PCA principal components or the ICA components of the gene expression profile, in four out of five cancer types that we tested. Conclusions Given high-dimensional “omics” data, the VAE is a powerful tool for obtaining a nonlinear low-dimensional embedding; it yields features that retain biological patterns that distinguish between different types of cancer and that enable more accurate tumor transcriptome-based prediction of response to chemotherapy than would be possible using the original data or their principal components.

Challenges and perspectives in use of artificial intelligence to support treatment recommendations in clinical oncology

“…Here, AI approaches were able to arrange tumor samples based on their RNA and tumor characteristics in a multidimensional space 32 as well as to predict response to chemotherapy based on tumor RNA. 33 Moreover, AI applications can be used to predict the survival of cancer patients, representing key information for any treatment decisions in oncology in order to adapt the treatment regime to achieve the highest possible survival outcome as well as quality of life. [34][35][36] By using data that can be extracted from most medical records, it has already been possible to train AIs that have improved predictions of overall survival for patients.…”

Artificial intelligence in pancreatic cancer

“…Wei et al used a variational autoencoder (VAE) algorithm to extract tumor transcriptome features. Regularized gradient boosted decision trees (XGBoost) were further used to predict chemotherapy drug response for cancer (for PC: AUROC 0.738; AUPRC 0.764) 80 . Cos et al collected preoperative activity metrics (step count, heart rate, and sleep time series) from patients with the help of wearable devices and built ML models to predict whether the pancreatectomy achieved the desired outcome (the absence of postoperative pancreatic fistulae, etc.).…”

Benchmarking Variational AutoEncoders on cancer transcriptomics data

“…Also, VAEs are proven to be useful in several applications, such as predicting drug response [14] and perturbation effects [15]. Using a semi-supervised approach and a VAE, Wei and Ramsey were able to predict response to chemotherapy for some cancer types [16].…”