Using FPKM for PECA/TimeReg

[Chloe-Thangavelu]

The manual says PECA can use TPM or FPKM as input but FPKM values are not usually used to compare between samples. Is it ok to use FPKM if we intend to use compare_diff or TimeReg, which compare between samples? (Also, for TimeReg, we have to take averages of the RNA replicates, can we do this with FPKMs? Or should we take the averages of the counts first then convert to FPKM?)

[durenzn]

For differential expression analysis, people did use count based method like DEseq2 and EdgeR for many years. However, those methods are worse than the simple ranksum test, which take FPKM or CPM as inputs, when the sample size is large. Please see a recent paper from Jingyi Jessica Li’s group: https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02648-4

Overall, to perform differential analysis, we just need to use the raw count and normalize it by total read counts, or FPKM/TPM after normalization would work.