Yishai Levin Ph.D., received his PhD at the University of Cambridge, UK. He then joined the Weizmann Institute to setup a proteomics core facility, which later became part of the Israel National Center for Personalised Medicine.
Today his work focuses on clinical proteomics in immunology and Alzheimer’s disease. In his spare time he likes off-road cycling and off-road driving.
How Skyline Saved Us From Publishing Erroneous DataOur story begins with a glycoproteomics project, with the aim of profiling glycopeptides from patient sera. We had two informatics tools at hand. One generates identifications, based on the MS/MS spectra, but not quantification (Byonic). The other, generates MS1 based, label free quantification from any list of peptide sequences (FlashLFQ). So we formatted the output from Byonic and analysed it using FlashLFQ to generate the quantitative data. After performing some Read More basic statistics, we ended up with a list of significant glycopeptides, and started to write the manuscript.
One thing kept bothering me about the process. We wrote an entire manuscript based on an automated quantitative output from a combination of two software tools, with no visualisation relating to how the peptide intensities were generated.
So I insisted we look at some of the significant peptides in Skyline, to make sure the quantification was correct.
We chose 9 glycopeptides and generated the MS1 based quantification in Skyline. We found that there was no correlation between the intensity values generated by Skyline and the output from FlashLFQ. This was very puzzling and after lengthy investigation, which I will discuss, we found the issue.
Once we fixed it, the correlation was >0.9 and we were confident out data is solid and worthy of publication.
[PDF] |