Transcripts per million ratio: applying distribution-aware normalisation over the popular TPM method

TL;DR

This paper introduces a novel RNA-seq normalisation method called Transcripts Per Million Ratio, which accounts for gene length and sample biases to improve differential expression analysis accuracy.

Contribution

The paper proposes a new normalisation technique that incorporates gene length and pan-sample biases, filling gaps in existing RNA-seq normalisation methods.

Findings

Improved accuracy in differential gene expression detection.

Better normalization across samples with varying gene lengths.

Addresses biases not considered in traditional TPM methods.

Abstract

Current popular methods in literature of RNA sequencing normalisation do not account for gene length when compared across samples, whilst adjusting for count biases in the data. This creates a gap in the normalisation as bigger genes in RNA sequencing accumulate more reads due to shotgun sequencing methods. As a result, the proportions of these reads inter-sample are not properly accounted for in current normalisation methods. Alternatively, methods which account for gene length do not account for the pan-sample biases in the data by accounting for a central read average. Thus, in order to fill in the gap in the literature, we propose a novel method of Transcripts Per Million Ratio and its relatives in RNA-sequencing differential expression normalisation that can be used in different conditions, which takes into account the gene length as well as relative expression in normalisation.