meaning of gene % prerank output

[idriskb]

Hello,

I am running prerank enrichment on my own ranking method. I am struggling to understand what does gene% mean in the output dataframe. I read in the documentation that:
tag %: Percent of gene set before running enrichment peak (ES),
gene %: Percent of gene list before running enrichment peak (ES),

But still I cannot understand it

[zqfang]

so, the

$$ \text{gene \%} = \text{(the position of the gene coresponding to enrichment score peak)} \div \text{(number of all genes in ranking list)}$$

and

$$ \text{tag \%} = \text{( number of leading genes)} \div \text{(number of genes in certain pathway that overalpped with the input ranking list)}$$

@Cher-HAN , Thanks for the correction (This metric is same to the GSEA_R)

[Cher-HAN]

Hi, zq.
The score used to rank in df_rnk is from 0 - 1. I draw a distribution boxplot and attched below.

[zqfang]

I don't have a clue what's going on here. Do you have duplicated gene names ? Can you set verbose=True or share the logging file ?

[Cher-HAN]

Thanks for the reply.
I am not familiar with "set verbose=True". Do you mean run df.info(verbose=True)? I ran this command and checked that no gene was repeated by set(). See pic. below.

I need to talk with my collaborators before sharing data, or I may make a sub-dataset and upload it later.

[Cher-HAN]

Hi zq, I realized I (maybe you also) misunderstand the meaning/calculation of gene %.
The gene% should be (the position of the gene coresponding to enrichment score peak) / (number of genes in the ranking list)
This formular is for gene set with positive enrichment score.

In my example, 18 leadings are at position 191 to 384 in the ranking list. And because this GO term has a negative enrichment score, the leading genes are located after the enrichment score peak. Thus the position of gene coresponding to the enrichment score peak is counted from end to start whcih is total number 389 - position 191 + 1,
gene % = (389 - 191 + 1)/389 = 51.16 %.

[zqfang]

Oooo.....h Yes. You'er right. The answer is updated now!