Integrated multi-omics analysis of ovarian cancer using variational autoencoders

Abstract

Cancer is a complex disease that deregulates cellular functions at various molecular levels (e.g., DNA, RNA, and proteins). Integrated multi-omics analysis of data from these levels is necessary to understand the aberrant cellular functions accountable for cancer and its development. In recent years, Deep Learning (DL) approaches have become a useful tool in integrated multi-omics analysis of cancer data. However, high dimensional multi-omics data are generally imbalanced with too many molecular features and relatively few patient samples. This imbalance makes a DL based integrated multi-omics analysis difficult. DL-based dimensionality reduction technique, including variational autoencoder (VAE), is a potential solution to balance high dimensional multi-omics data. However, there are few VAE-based integrated multi-omics analyses, and they are limited to pancancer. In this work, we did an integrated multi-omics analysis of ovarian cancer using the compressed features learned through VAE and an improved version of VAE, namely Maximum Mean Discrepancy VAE (MMD-VAE). First, we designed and developed a DL architecture for VAE and MMD-VAE. Then we used the architecture for mono-omics, integrated di-omics and tri-omics data analysis of ovarian cancer through cancer samples identification, molecular subtypes clustering and classification, and survival analysis. The results show that MMD-VAE and VAE-based compressed features can respectively classify the transcriptional subtypes of the TCGA datasets with an accuracy in the range of 93.2-95.5% and 87.1-95.7%. Also, survival analysis results show that VAE and MMD-VAE based compressed representation of omics data can be used in cancer prognosis. Based on the results, we can conclude that (i) VAE and MMD-VAE outperform existing dimensionality reduction techniques, (ii) integrated multi-omics analyses perform better or similar compared to their mono-omics counterparts, and (iii) MMD-VAE performs better than VAE in most omics dataset.

Let-7i Reduces Aggressive Phenotype and Induces BRCAness in Ovarian Cancer Cells

“…Research on the molecular characterization of ovarian cancer on genomic, proteomic, and other levels has been ongoing for over a decade, launched by the flagship The Cancer Genome Atlas (TCGA) project [72]. Various scientific efforts have resulted in characterization of chromosomal aberrations, genomic rearrangements, and signaling pathway disruptions, as well as post-translational modifications [73,74]. Clinically, this has resulted in the development of tumor agnostic clinical trials, i.e., KEYNOTE trials [75,76].…”

“…Combination of genomics, epigenomics, transcriptomics, and proteomics will provide us with a more complete picture of the disease progress and the treatment options most likely to succeed. Experts from multiple disciplines of medicine and research working together to treat patients will facilitate progress and enhance patient outcomes [17,73,77,78].…”

Multi-omics and deep learning provide a multifaceted view of cancer

“…The most commonly used deep neural network architectures for latent factor modeling of multi-omics datasets are autoencoders. Different variants of auto-encoder architectures have been used for the analysis of multi-omics datasets [23,25,26]. However, these studies have demonstrated the usefulness of the methods on only selected cancer types and have not extensively and comprehensively tested their methods on a variety of use cases.…”

XOmiVAE: an interpretable deep learning model for cancer classification using high-dimensional omics data

“…Variational autoencoder (VAE) [ 17 ] is one of the emerging deep learning methods that have shown promise in embedding omics data to lower-dimensional latent space. With a classification downstream network, the VAE-based model is able to classify tumour samples and outperform other machine learning and deep learning methods [ 2 , 15 , 45 , 46 ]. Among them, OmiVAE [ 46 ] is one of the first VAE-based multi-omics deep learning models for dimensionality reduction and tumour type classification.…”

“…Similar to OmiVAE, DeePathology [ 2 ] applied two types of deep autoencoders, contractive autoencoder and VAE, with only the gene expression data from the GDC dataset, and reached accuracy of 95.2% for the same tumour type classification task. Hira et al [ 15 ] adopted the architecture of OmiVAE with maximum mean discrepancy VAE and classified the molecular subtypes of ovarian cancer with an accuracy of 93.2–95.5%. Zhang et al [ 45 ] synthesized previous models and developed a unified multi-task multi-omics deep learning framework named OmiEmbed, which supported dimensionality reduction, multi-omics integration, tumour type classification, phenotypic feature reconstruction and survival prediction.…”