remove the unused PosteriorErrorProbability field from the FlashLFQ.Identification class
removed the unused RtPeakHypothesis, RtStdDev, and RtIqrRange properties from the chromatographic peak class
Previously, the MBR score was unused by FlashLFQ. As of this PR, the MBR score is still unused by FlashLFQ. Subsequent PRs will use the MBR score for calculation of FDR.
This PR alters how MBR works through changes to retention time prediction. Previously, we used a local alignment window of one minute to predict the retention time of MBR acceptors. Now, instead of setting a time for the alignment window, we use the 4 closest anchor peptides (peptides MS2-detected in both donor and acceptor runs), two on either side of the donor peptide, to predict the retention time for all MBR acceptors.
I tested this version of FlashLFQ using the FlashLFQ E. coli spike in vignette data. In that data, two samples are compared. One contains e. coli proteins spiked into a human sample @ 1x concentration, the other contains e. coli proteins spiked into a human sample @ 2x concentration. The table below shows the results of FlashLFQ's protein quantification of the ecoli proteins for the current release and incorporating the changes in this PR.
The retention time prediction method used by MBR was changed, which changes the results of MBR. This change explains the results shown below.
First part of my MBR overhaul.
Previously, the MBR score was unused by FlashLFQ. As of this PR, the MBR score is still unused by FlashLFQ. Subsequent PRs will use the MBR score for calculation of FDR.
This PR alters how MBR works through changes to retention time prediction. Previously, we used a local alignment window of one minute to predict the retention time of MBR acceptors. Now, instead of setting a time for the alignment window, we use the 4 closest anchor peptides (peptides MS2-detected in both donor and acceptor runs), two on either side of the donor peptide, to predict the retention time for all MBR acceptors.
I tested this version of FlashLFQ using the FlashLFQ E. coli spike in vignette data. In that data, two samples are compared. One contains e. coli proteins spiked into a human sample @ 1x concentration, the other contains e. coli proteins spiked into a human sample @ 2x concentration. The table below shows the results of FlashLFQ's protein quantification of the ecoli proteins for the current release and incorporating the changes in this PR.
The retention time prediction method used by MBR was changed, which changes the results of MBR. This change explains the results shown below.
<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns="http://www.w3.org/TR/REC-html40">
| Current Release | Scoring Update (2 Anchors | Scoring Update (3 anchors, this PR) -- | -- | -- | -- No. E.Coli Proteins Quantified in both | 515 | 522 | 524 2x:1x mean protein intensity ratio | 1.999661 | 1.947146 | 1.953307 2x:1x Std.Dev protein intensity ratio | 0.769428 | 0.707974 | 0.719638