Identification of expression signatures for non-small-cell lung carcinoma subtype classification

• DOI: 10.1093/bioinformatics/btz557
• Publication Date: 2019-07-09T11:11:18Z
• Authors: Ran Su; Jiahang Zhang; Xiaofeng Liu; Leyi Wei
• Keywords: Discriminative model; Identification

Non-small-cell lung carcinoma (NSCLC) mainly consists of two subtypes: squamous cell (LUSC) and adenocarcinoma (LUAD). It has been reported that the genetic epigenetic profiles vary strikingly between LUAD LUSC in process tumorigenesis development. Efficient precise treatment can be made if subtypes identified correctly. Identification discriminative expression signatures explored recently to aid classification NSCLC subtypes.In this study, we designed a model integrating both mRNA long non-coding RNA (lncRNA) data effectively classify NSCLC. A gene selection algorithm, named WGRFE, was proposed identify most within recursive feature elimination (RFE) framework. GeneRank scores considering level correlation, together with importance generated by classifiers were all taken into account improve performance. Moreover, module-based initial filtering genes performed reduce computation cost RFE. We validated algorithm on The Cancer Genome Atlas (TCGA) dataset. results demonstrate developed approach small number for accurate subtype particularly, here first time show potential role LncRNA building computational models.The R implementation is available at https://github.com/RanSuLab/NSCLC-subtype-classification.