Hi Jordane,
I see this error also, skyline assigns the wrong peak even when there is plenty of info (RT, dotp, intensity, shape) suggesting that another peak is better, usually happens when the retention time is of but in some cases it can be pretty good and still give this result (attached are a couple of examples with 5600 data in DIA). For me it happens with precursor and DIA based quant. The only way I have been able to remove this is to select the correct peak manually, I did this for 2000 peptides it took me about a day and half to get through them all. However there were still errors and when I checked some of my peptides with signifcant differences they had this error.

I guess I just wanted to add support to your comment, to make sure that the developers know this is not a isolated problem and one that seems like it should be relatively easy to fix in most cases.

cheers
Rob

Hi Rob and Jordane,
Believe me, I am aware of this issue, and it is currently my top priority. I am extremely hopeful that I will be able to improve peak picking for this kind of data by next release. Until the, unfortunately, you will have to do what Rob reports doing and correct these types of errors manually. The Skyline replicate comparison graphs and synchronized zooming are reasonably powerful tools for achieving this manual correction, though I certainly agree I need to reduce the amount it is needed.

To do that, I intend to integrate the machine learning methodology outlined in the mProphet paper:

http://www.ncbi.nlm.nih.gov/pubmed/21423193

With some added features/scores and different decoy model for DIA data.

It should work considerably better than what Skyline uses today. My apologies for not getting this integrated sooner.

Thanks for your feedback and examples.

--Brendan

Hey Brendan,
Thanks for getting back, it's good to know that you know about this issue I look forward to trying the update!
cheers
Rob

Hi Rob and Brendan,

Thank you very much for your responses. I was afraid I was doing something wrong. And even with these mistakes, skyline is really useful!!
Jordane

Hi Jordane, Rob and Brendan,
I am a new user of Skyline, and not very experienced in MSpec, and would love to have Skyline give peak area quant data so that protein fold changes could be calculated across samples.
I've also had issues with Skyline assigning the wrong peak across a set of replicate samples, especially considering that the treated samples will have peptides/proteins that are absent in the untreated samples.

The data was generated from a 5600 and I used both Mascot and ProteinPilot to search the wiff files. I noticed in a handful of peptides that the assignment across my replicates was slightly better when group.xml files rather than mgf files were used to create the library. I tried a similar scan time limitation in the parameters but that didn't seem to matter. I have found it very messy to correct the assignments.

Skyline seemed to have difficulty in sticking close to the expected RT for a peptide (especially if that peptide peak is absent in the sample) and instead selected a peak at an RT for another peptide of similar mass. That was confirmed by searching the protein pilot results for peptides of a similar mass and at the observed RT. My question is if Skyline has incorrectly assigned a peak that belongs to another peptide does that mean that peak is unable to be reassigned to the correct peptide? Am I able to correct the peak assignment to the correct peptide even if the peptide does not appear in the peptide tree?

I hope I've been clear enough.
Thanks for you help
Connie