|Title||»||Improve BlibFilter best spectrum picking for targeted MS/MS data|
Kaipo has been working on this, but I guess it isn't yet represented in the issues list. For targeted MS/MS data, we recognize that the current BlibFilter algorithm for picking the best spectrum frequently ends up picking a much lower signal-to-noise spectrum, because the spectra are all normalized and scored using a dot-product. So that when there are many low signal-to-noise spectra, their noise will end up making them score highest dot-products with each other.
Obviously, we need to fix this to an algorithm that more frequently picks a spectrum closer to the apex of the peptide elution curve.