How to set a cutoff TPM value to determine whether a gene is expressed or not?

Hi, I am working on a bioinformatics project where I need to set a threshold to distinguish between expressed and unexpressed genes.

Here is the percentile distribution of TPM values in my dataset:

0th:    0.0
10th:   0.0
25th:   0.0
33rd:   0.0
50th:   0.0
67th:   0.0716
70th:   0.1156
75th:   0.2608
90th:   5.163
100th:  61032.6

I couldn’t find any standard cutoff value. Should I select 1 as the threshold point?

value tpm expression gene

• 58 views • link

updated 3 days ago by ATpoint

• written 3 days ago by Muhammad Tahmid •

1 answer

This was asked many times before, and the answer is that there is no cutoff since biology is continuous. Often people try to find inflexion points in ranked expression to at least have a non-arbitrary cutoff, but that is not biology-driven, so just using 1 might just be as good (or bad). RNA-seq is inherently relative, so statements on expression TRUE/FALSE is anyway not what it is made for. Since in comparative analysis you need decent counts to have power, I often just use filterByExpr() from edgeR and continue with everything it retains.

• 0 views • link

written 3 days ago by ATpoint