Min-redundancy and max-relevance multi-view feature selection for predicting ovarian cancer survival using multi-omics data

Yasser El-Manzalawy,Tsung-Yu Hsieh,Vasant Honavar,Manu Shivakumar,Dokyoon Kim

doi:10.1186/s12920-018-0388-0

Yasser El-Manzalawy, Tsung-Yu Hsieh + Show 3 more

Open Access

https://doi.org/10.1186/s12920-018-0388-0

Copy DOI

Abstract

BackgroundLarge-scale collaborative precision medicine initiatives (e.g., The Cancer Genome Atlas (TCGA)) are yielding rich multi-omics data. Integrative analyses of the resulting multi-omics data, such as somatic mutation, copy number alteration (CNA), DNA methylation, miRNA, gene expression, and protein expression, offer tantalizing possibilities for realizing the promise and potential of precision medicine in cancer prevention, diagnosis, and treatment by substantially improving our understanding of underlying mechanisms as well as the discovery of novel biomarkers for different types of cancers. However, such analyses present a number of challenges, including heterogeneity, and high-dimensionality of omics data.MethodsWe propose a novel framework for multi-omics data integration using multi-view feature selection. We introduce a novel multi-view feature selection algorithm, MRMR-mv, an adaptation of the well-known Min-Redundancy and Maximum-Relevance (MRMR) single-view feature selection algorithm to the multi-view setting.ResultsWe report results of experiments using an ovarian cancer multi-omics dataset derived from the TCGA database on the task of predicting ovarian cancer survival. Our results suggest that multi-view models outperform both view-specific models (i.e., models trained and tested using a single type of omics data) and models based on two baseline data fusion methods.ConclusionsOur results demonstrate the potential of multi-view feature selection in integrative analyses and predictive modeling from multi-omics data.

Highlights

Large-scale collaborative precision medicine initiatives (e.g., The Cancer Genome Atlas (TCGA)) are yielding rich multi-omics data
Single-view models for predicting ovarian cancer survival We evaluated Random Forest (RF), eXtreme Gradient Boosting (XGB), and Logistic Regression (LR) classifiers trained using each of the individual views with the top k features selected using Lasso feature selection algorithm for choices of k = 10,20,30, ..., 100.Tables 3, 4 and 5 report the performance of the resulting classifiers averaged over 10 different 5-fold cross-validation experiments
It should be noted that when the performance of single-view models is estimated using a single 5-fold cross-validation experiment, the best observed area under ROC curve (AUC) scores were 0.70, 0.55, and 0.69 for models built from the copy number alteration (CNA), methylation, and RNA-Seq views, respectively

Summary

Introduction

Large-scale collaborative precision medicine initiatives (e.g., The Cancer Genome Atlas (TCGA)) are yielding rich multi-omics data. Integrative analyses of the resulting multi-omics data, such as somatic mutation, copy number alteration (CNA), DNA methylation, miRNA, gene expression, and protein expression, offer tantalizing possibilities for realizing the promise and potential of precision medicine in cancer prevention, diagnosis, and treatment by substantially improving our understanding of underlying mechanisms as well as the discovery of novel biomarkers for different types of cancers. Such analyses present a number of challenges, including heterogeneity, and high-dimensionality of omics data. The resulting methods have been successfully used to predict the molecular abnormalities that impact both clinical outcomes and therapeutic targets [5, 10, 12,13,14,15,16]

Methods

Results

Discussion

Conclusion