TPM in RNA-Seq: What It Is & How It Differs from FPKM

Raw read counts from RNA-Seq analysis cannot be directly compared across genes or samples.

This is because longer genes naturally accumulate more mapped reads, and samples with a higher total number of sequenced reads will have more reads mapped to every gene.

To address these biases, various normalization methods have been proposed. This page explains TPM. In the past, FPKM/RPKM was widely used, but it has since been criticized for not accurately representing gene expression levels. As a result, TPM has become the preferred metric.

TPM stands for 'Transcripts Per Million' and was proposed as an alternative to FPKM/RPKM.

Like FPKM/RPKM, TPM normalizes for both the total number of mapped reads (scaled to one million) and transcript length (scaled to 1,000 bases). The key difference is the order of operations: TPM first corrects for transcript length, then normalizes by the total read count.

The formula is as follows, where \(q_i\) is the number of mapped reads and \(l_i\) is the transcript length:

TPM can also be expressed in terms of FPKM as follows:

The exact TPM calculation can vary slightly between software tools. For instance, some tools use the effective length rather than the actual transcript length for \(l_i\).

The effective length is calculated as follows:

\(μ_{FLD}\) represents the average fragment length.

Using the effective length in TPM calculations is considered to provide a more accurate correction for length bias.