A robust fuzzy rule based integrative feature selection strategy for gene expression data in TCGA

• DOI: 10.1186/s12920-018-0451-x
• Publication Date: 2019-01-31T13:05:29Z
• Authors: Shicai Fan; Jianxiong Tang; Qi Tian; Chunguo Wu
• Keywords: Gene signature

Lots of researches have been conducted in the selection gene signatures that could distinguish cancer patients from normal. However, it is still an open question on how to extract robust features.In this work, a signature strategy for TCGA data was proposed by integrating expression data, methylation and prior knowledge about biomarkers. Different traditional integration method, expanded 450 K were applied instead original array reported biomarkers weighted feature selection. Fuzzy rule based classification method cross validation model construction performance evaluation.Our selected features showed prediction accuracy close 100% with fuzzy 6 cancers TCGA. The our similar other integrative models or RNA-seq only model, while independent obviously better than 5 models. extracted more robust, had potential get results.The results indicated would cover genes, greater capacity retrieve genes compared data. Also, promising way improve performance. PTCHD3 as discriminating 3 out cancers, which suggested might play important role risk be worthy intensive investigation.