Biomarker Identification in Colorectal Cancer Using Subnetwork Analysis with Feature Selection

Conference proceedings article

Authors/Editors

Strategic Research Themes

Publication Details

Author list: Kozuevanich S., Meechai A., Chan J.H.

Publisher: Springer

Publication year: 2020

Volume number: 1149 AISC

Start page: 119

End page: 127

Number of pages: 9

ISBN: 9783030440435

ISSN: 2194-5357

eISSN: 2194-5357

URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85083645970&doi=10.1007%2f978-3-030-44044-2_12&partnerID=40&md5=9943a5380c47d534e56e5c44114a9e24

Languages: English-Great Britain (EN-GB)

View on publisher site

Abstract

Gene Sub-Network-based Feature Selection (GSNFS) is an efficient method for handling case-control and multiclass studies for gene sub-network biomarker identification by an integrated analysis of gene expression, gene-set and network data. However, GSNFS has produce considerably high number of sub-network and has not assessed the importance of each sub-network. Recently, we have incorporated 2 feature selection techniques; correlation-based and information gain into the GSNFS workflow to help reduce the number and assess the importance of each individual sub-network. The extended GSNFS method was clearly shown to identify good candidate gene subnetwork markers in lung cancer. In this work, we applied a similar work flow to colorectal cancer. First, the top- and bottom- 5 ranked gene-sets were selected and investigated the classification performance. Similarly, the top-ranked gene-sets showed a better performance than the bottom-ranked gene-sets. The identified top-ranked gene-sets such as TNF-beta and MAPK signaling pathway were known to relate to cancer. In addition, the characteristic of top identified pathway network was further analyzed and visualized. SMAD3, a gene that was reported to be related to cancer by many studies, was mostly found to have the highest neighbor in 4 datasets. The results in this study has confirmed that GSNFS combined with feature selection is very promising as significantly fewer subnetworks were needed to build a classifier and gave a comparable performance to a full dataset classifier. © 2020, The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

colorectal cancer, Correlation-based feature selection, gene expression analysis, gene-set, Information Gain feature selection