P value cutoff selection for functional analysis

MetaboAnalyst 6.0: While using the functional analysis tool for MS peaks, the p value cutoff is at 10% top peaks by default. Is this automatically calculated depending on my dataset? Should I always prefer to select the 10% top peaks as p value cutoff or should I set the p-value cutoff as 0.05 significance level? Are there some suggestions for the cutoff value selection for the functional analysis?

P-values, t-statistics or fold changes are commonly used to help select peaks that are changed or “perturbed” in functional analysis. There are two main considerations here:

function is a group behaviour. For instance, a pathways involve ~40 compounds could map to ~200 peaks. If we would like to have a good understanding of functions based on peaks, an empirical rule is 300~500 peaks for detecting different functions.

background is the universe against which the enrichment is calculated. The assumption is that the majority (say, 90%) of the metabolome will remain stable, so that we can use random sampling (sample size the same as the significant list) from this universe to compute null distribution.

From the above estimation, the algorithm (i.e. mummichog) need ~5000 peaks as universe, 300~500 peaks as significant peaks to work robustly. This is usually the case in LC-MS untargeted metabolomics based on Orbitrap or TOF.

This topic was automatically closed after 3 days. New replies are no longer allowed.