why my volcano plot looks strange?

Question

[hellocita]

Hi all, I am dealing with an LC-MS Metabolomics data. I perform log transform, quantile normalization and use linear model to regress out batch. And use wilcox test to figure out which metabolite is significantly changed. However, the pvalue and log2 fold change of metabolites are weird. There are lots of dots near the middle line. Please help and give me some advice!

[cpad0112]

is it log10p or -log10p on y-axis?

Answer 1

[swbarnes2]

Reverse and log scale the y axis for starters.

[hellocita]

Thanks for your help! Now it looks better, and there are still many dots near the middle line, was this acceptable?

[ATpoint]

First thing to do would be to not use custom statistics but a proper framework such as limma. There are quite some peptides (or whatever you measure) with tiny effect sizes but still small pvalues. Is this corrected for multiple testing?

[hellocita]

The smallest p value is 0.02, so I did not correct it

[ATpoint]

Then your analysis is basically meaningless, FDR is not optional, it always needs to be applied to avoid a large amount of false-positives. I strongly recommend to either work with a statistician or use dedicated statistical software such as limma which does all that under the hood.