This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Which one? Log2foldchange or t statistic

Hello , When I prepare my gene_list for gseGo() I tried Log2foldChange or t statistic ,separately , to order the list. However, I got different results for each. Which one do you recommend?

I used this code after deseq2 to get the differentially expressed gene list:

res <- results(dds, contrast=c("group", "KO", "wt")).

Afterwards, I just created the list of genes using either "Lof2foldChange" or "t statistic" and the gene names as the row names.

Thanks for your help in advance.

gsego
• 2,755 views • link
updated by nn • nn • written by Fatemeh • Fatemeh

Definitely, you should not use logFC to rank the genes for gene set enrichment analysis. So the answer is 't-statistics' (or "Wald-statistics" in DESeq2 output ). If you want to use logFC consider this line of code to convert logFC to a metric that can take into account both statistical significance and effect size:

metric = -log10(pvalue)/sign(log2FC)
• 0 views • link
written by Hamid Ghaedi Hamid Ghaedi

Thanks very much Hamid for your reply. It was a big help.

• 0 views • link
written by Fatemeh • Fatemeh

1 answer

https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-017-1674-0

I usually go with t statistics

• 0 views • link
written by nn • nn

Thanks a lot for your answer and provided link.

• 0 views • link
written by Fatemeh • Fatemeh

The t statistic does NOT blend significance and effect size. It's significance only.

• 0 views • link
written by dsull dsull

You're right, I removed my mistake.

• 0 views • link
written by nn • nn

Log in to answer this question.

More posts like this