Flaver: mining transcription factors in genome-wide transcriptome profiling data using weighted rank correlation statistics

Abstract

AbstractBackgroundMining key transcription factors (TFs) in genome-wide transcriptome profiling data has been an active research area for many years and it has been partially solved by mathematically modelling the ranking orders of genes in the target gene-set for the TF of interest in the gene-list ranked by expression values, called gene-set enrichment analysis (GSEA). However, in some application scenarios the gene-set itself also has a rank attribute, such as the putative target gene-set predicted by the Grit software and other alternatives like FIMO and Pscan. New algorithms must be developed to analyze these data properly.Methodology/Principal FindingsBy implementing the weighted Kendall’s tau statistic, we proposed a method for genome-wide transcriptome profiling data mining that can identify the key TFs orchestrating a profile. Theoretical properties of the proposed method were established, and its advantages over the GSEA approach were demonstrated when analyzing the RNA-Atlas datasets. The results showed that the top-rated TFs by our method always have experimentally supported evidences in the literatures. Benchmarking using gene ontology (GO) annotations in the AmiGO database indicated that the geometry performance (SQR_P) of our method is higher than GSEA in more than 14% of the cases.SignificanceThe developed method is suitable for analyzing the significance of overrepresentation of ranked gene-sets in a ranked gene-list. A software implementing the method, called “Flaver”, was developed and is publicly available at http://www.thua45.cn/flaver under an academic free license.Author SummaryIdentification of the regulation roles of TFs in the transcriptome is fundamental in understanding various biological processes. Improve the performance of the prediction tools is important because accurate TF-mining in transcriptome data can finely improve the efficiency of wet-lab experiments. Also, genome wide TF-mining can provide new target genes for transcriptome regulation analysis in system biology perspective. This study developed a new TF-mining tool based on weighted rank correlation statistical method. The tool has better performance in analyzing ranked gene-set and ranked gene-list than its competitor, the GSEA tool. It can help the researchers in identification of the most important TFs in transcriptional data.