Standard Fdr Threshold In Chip-Seq & How To Co-Relate It To Q-Value [Macs14]

Question

[Sukhi Singh]

Hi,

My first question is, what is the standard measure of FDR (if any) while calling peaks in ChIP-Seq.

And, how to co-relate the emperical FDR from Macs14 (peak caller) to a q-value threshold. Is it just, empFDR=(q-value/10)*100 or one cannot make such remarks.

Can we modulate this threshold on the case to case basis. There might be a case, when your protein is low expressed, so the enrichment is low. And when you call peaks, they pass the p-value threshold, but the enrichment at most postions is same or less than the control, so FDR is high. What to do in that case.

Cheers

P.S. The empirical FDR in Macs 14 is defined as Number of control peaks / Number of ChIP peaks.

[Arun]

q-value is FDR corrected p-value.

[Sukhi Singh]

yeah I agree and this led me to read this post about p,FDR and q values and these lines say the thing

a p-value of 0.05 implies that 5% of all tests will result in false positives. An FDR adjusted p-value (or q-value) of 0.05 implies that 5% of significant tests will result in false positives. The latter is clearly a far smaller quantity.

Answer 1

[Sean Davis]

The false discovery rate is just that. There is no obvious cutoff for such a value, but the biological interpretation is meant to be relatively clear. If a "cutoff" of 0.05 is used, 95% of the results are not false discoveries. Following that reasoning, any cutoff <1 is justifiable, but it may be impractical.

[Sukhi Singh]

Thanks Sean, so one can say that a protein is low expressed and using a threshold of 0.5 (50% chance the peaks might be wrong) will be fine