• Sign In
  1. Requests

DotP score calculation

support

View Request

view list print
DotP score calculation YZ  2015-03-06 08:54
 
Hello,

I am always wondering how exactly dotP score is calculated in skyline. Although I have seen you answered this question before, I am still not sure about the right function. I have seen the paper http://www.ncbi.nlm.nih.gov/pubmed/24623587, the normalized spectral contrast angle seems the picked scoring function, which I have taken a screenshot. Is this the one that skyline uses? your paper http://www.ncbi.nlm.nih.gov/pubmed/19326923 used Costa-Soares correlation score to determine the similarity between library spectra and acquire msms. how does that related to dotP score?

The other question is how bibliospec picks a best spectrum. You answered previously, "based on sum of dot products between it and all other spectra for the same peptide". Since I can't understand how dotP is calculated, so I couldn't understand the principle. Could you explain a little bit more?

Thank you very much!!

Best,
Yiying
 
 
Brendan MacLean responded:  2015-03-06 12:04
Hi Yiying,
Yes, everything in Skyline that uses "dotp" (dotp, idotp, rdotp) is actually a "normalized contrast angle". Thanks for including a link to the Toprak, MCP 2014 paper.

Despite the fact that I was an author on the Prakash, J. Proteome Res. 2009 paper for which you have also provided a link. I was not involved in choosing the Costa-Soares rank correlation score for that paper. That decision actually precedes the first addition of a spectrum correlation function to Skyline, and I was never able to convince myself it was a good enough idea to use it in Skyline.

Originally, Skyline used a standard dot product (http://en.wikipedia.org/wiki/Dot_product), as this was what was in use in BiblioSpec, the spectral library search and build tool developed at the MacCoss lab (Frewen, Curr. Protoc. Bioinformatics 2007 - http://www.ncbi.nlm.nih.gov/pubmed/18428681). However, after meeting with Toprak and Gillet at ETH in July 2013, they convinced me that for the smaller numbers of fragments usually seen in targeted experiments, the normalized contrast angle was better. I had already noted that the cos(angle) dot-product values were not doing a good job of distinguishing good from poor within Skyline.

For spectral library building, we have changed what BiblioSpec does to pick a best spectrum from what was described in the original Frewen 2007 paper, and which I described previously, and you quote. That algorithm did not perform well on searched PRM data, and had a tendency to pick very noisy, low intensity spectra for that data type. We now use a combination of best search engine score and highest TIC within the spectrum to pick, and this does much better at picking the apex spectrum in PRM.

Thanks for posting your questions to the support board. I hope this response proves useful to you. Good luck using Skyline in your research.

--Brendan
 
TY responded:  2025-02-05 21:32
Hi Brendan,

I was just searching for information about dotp in Skyline, and this thread was really helpful!

I have a couple of questions regarding the implementation of spectrum preprocessing and the calculation of dotp (equivalent to the normalized contrast angle).
What would be the easiest way to perform these tasks? I know there are some resources available for the Skyline command line interface and its source code.
I assume there might be a suitable commandlet in these resources.
If you’re available, I would appreciate any advice you could provide on this topic.

Best regards,
TY