Adding an RT correction based on peak groups wouldn't be that hard... right?

[wkumler]

While I don't understand retention time correction algorithms well enough to implement something like Obiwarp, a basic version using peak groups doesn't seem that hard. Basically I'd need the RTs for a couple peaks on a per-file basis and then apply a linear (loess?) interpolation that lines those peaks up at the same RT.

Option 1 would be integrating a couple peaks per-file which is a painful process. Option 2 would be applying some kind of heuristic like "match up the maximum intensity points" (maximize correlation?) within a large RT range. I know I might actually use the second one and would probably never do the first (except maybe on a subset?).

You could even pick peaks based on the BPC/TIC for a couple high-intensity peaks if they were consistent across files, or apply it to a single chromatogram. That would also likely get a better spread of RTs for interpolation and avoid the temptation to use six different internal standards that all elute within minutes 5-6.

Also, this could be done on a subset of the files (e.g. our Pooled samples) and then interpolated between using timestamp in the metadata.

Expected outputs would be a data frame of filename, init_rt, and corr_rt that could be joined back to the original MS1 data. Inputs would have to be like feature, filename, and peak_rt.