Cancer Characteristic Gene Selection via Sample Learning Based on Deep Sparse Filtering

Jian Liu,Xuesong Wang,Z Jane Wang,Yuhu Cheng,Lin Zhang

doi:10.1038/s41598-018-26666-0

Abstract

Identification of characteristic genes associated with specific biological processes of different cancers could provide insights into the underlying cancer genetics and cancer prognostic assessment. It is of critical importance to select such characteristic genes effectively. In this paper, a novel unsupervised characteristic gene selection method based on sample learning and sparse filtering, Sample Learning based on Deep Sparse Filtering (SLDSF), is proposed. With sample learning, the proposed SLDSF can better represent the gene expression level by the transformed sample space. Most unsupervised characteristic gene selection methods did not consider deep structures, while a multilayer structure may learn more meaningful representations than a single layer, therefore deep sparse filtering is investigated here to implement sample learning in the proposed SLDSF. Experimental studies on several microarray and RNA-Seq datasets demonstrate that the proposed SLDSF is more effective than several representative characteristic gene selection methods (e.g., RGNMF, GNMF, RPCA and PMD) for selecting cancer characteristic genes.

Highlights

The advances of DNA microarray and deep sequencing technologies have made it possible for biologists to measure expression levels of thousands of genes simultaneously[4,5]
Since gene expression datasets generally are with high dimensional features and small sample size, stacked denoising autoencoder (SDAE) and deep belief networks (DBNs) suffer from serious overfitting when applied to gene expression data
We test our method on two RNA-Seq datasets, i.e., esophageal cancer (ESCA) and squamous cell carcinoma of head and neck (HNSC)

Summary

Introduction

The advances of DNA microarray and deep sequencing technologies have made it possible for biologists to measure expression levels of thousands of genes simultaneously[4,5]. The most commonly used models include deep belief networks (DBNs)[8], stacked auto-encoders (SAEs)[10], and convolutional neural networks (CNNs)[11] These models have been successfully applied to numerous fields (e.g., image processing, natural language processing, and medical data analytics) and achieved promising performances. The major steps of the approach are described as follows: (1) Use DBN to extract high level representations of the gene expression profiles; (2) Apply a feature selection method to rank genes; (3) Obtain the selected genes using active learning. Both SDAE15 and DBN16 are supervised methods, and can learn high level features of the gene expression data. Sample learning transforms the sample space of gene expression data and ensures that the features (or genes) can be better represented by the transformed sample space so that we can specify the exact characteristic genes from the transformed sample space

Objectives

Methods

Results

Conclusion