Identifying common transcriptome signatures of cancer by interpreting deep learning models

Abstract

Cancer is a set of diseases characterized by unchecked cell proliferation and invasion of surrounding tissues. The many genes that have been genetically associated with cancer or shown to directly contribute to oncogenesis vary widely between tumor types, but common gene signatures that relate to core cancer pathways have also been identified, signifying that cancer cases display common hallmark molecular features. It is not clear however whether there exist additional sets of genes or transcriptomic features that are less well known in cancer biology but that are also commonly deregulated across several cancer types. Here, in order to agnostically identify transcriptomic features that are commonly shared between cancer types, we used RNA-Seq datasets encompassing thousands of samples from 19 healthy tissue types and 18 solid tumor types to train three feed-forward neural networks, based either on protein-coding gene expression, lncRNA expression or splice junction use, to distinguish between healthy and tumor samples. All three models achieve high precision, recall and accuracy on test sets derived from 13 datasets used during training and on an independent test dataset, indicating that our models recognize transcriptome signatures that are consistent across tumors. Analysis of attribution values extracted from our models reveals that genes that are commonly altered in cancer by expression or splicing variations are under strong evolutionary and selective constraints, suggesting that they have important cellular functions. Importantly, we found that genes composing our cancer transcriptome signatures are not frequently affected by mutations or genomic alterations and that their functions differ widely from the genes genetically associated with cancer. Finally, our results also highlighted that deregulation of RNA-processing genes and aberrant splicing are pervasive features across a large array of solid tumor types. The transcriptomic features that we highlight here define cancer signatures that may reflect causal variations or consequences of disease state, or a combination of both.