RLeave: an in silico cross-validation protocol for transcript differential expression analysis

TL;DR

RLeave is a novel in silico validation protocol combining traditional differential expression analysis with machine learning techniques to identify key transcripts for further validation in RNA-Seq studies.

Contribution

It introduces the RLeave algorithm that integrates edgeR, leave-one-out sampling, and decision trees for improved transcript relevance detection.

Findings

Confirmed key miRNAs in diabetes and leukemia with RT-qPCR

Effective in highlighting transcripts for validation in RNA-Seq

Applicable for in silico and in vitro validation processes

Abstract

Background and Objective: The massive parallel sequencing technology facilitates new discoveries in terms of transcript differential analysis; however, all the new findings must be validated, since the diversity of transcript expression may impair the identification of the most relevant ones. Methods: The proposed RLeave algorithm (implemented in the R environment) utilizes a combination of conventional analysis (classic edgeR) together with other mathematical methods (Leave-one-out sample technique and Decision Trees validation) to identify more relevant candidates to be in vitro or in silico validated. Results: The RLeave protocol was tested using miRNome expression analysis of two sample groups (diabetes mellitus and acute lymphoblastic leukemia), and both had their most important differentially expressed miRNA confirmed by RT-qPCR. Conclusion: This protocol is applicable in RNA-SEQ research, highlighting the most relevant transcripts for in silico and/or in vitro validation.