Quantifying the ~75-95% of Peptides in DIA-MS Datasets that were not Previously Quantified
We demonstrate an algorithm termed GoldenHaystack (GH) that, compared to the leading DIA-MS algorithm, (a) quantifies and identifies with better FDR accuracy the peptides found in FASTA search spaces ([~]5-25% of analytes in DIA-MS datasets), (b) quantifies the remaining [~]75-95% of analytes that were previously unquantified, and (c) runs [~]40-200x faster (or [~]1-10x faster than the LC-MS). Specifically, without a FASTA or spectral library, GH can deconvolute and accurately quantify chimeric LC-MS spectra. The central idea that enables this claim is: for sufficiently sized projects (e.g., [≥] [~]50 LC-MS files), pairs of peptides that co-elute in one subset of LC-MS files do not exactly co-elute in a different subset of files. GH thus analyzes a project holistically: it uses multi-partite matching to match fragment ions across all samples, separates and regroups the fragment ions into unique analyte signatures, reduces stochastic noise, and then quantifies those unique analyte signatures.