Basic-peptide depletion and precursor masking in bottom-up proteomics
Biomarker proteomics studies report protein identities at the canonical precursor level, yet the underlying mass-spectrometry evidence consists of peptides whose physicochemical properties may diverge from the annotated protein. We reanalyzed peptide-level evidence from public datasets with a proteoform-aware pipeline that maps detected peptides to canonical sequences under a five-level confidence taxonomy, computes sequence-derived physicochemical features for precursors, peptides, and detected regions, and tests class-composition bias against digestion-matched backgrounds. Across ten public studies spanning five result-file formats and 70,020 peptide-evidence rows, detected peptides are depleted in the basic isoelectric-point class relative to their own proteins' theoretical tryptic peptides (fixed-effect odds ratio 0.45, 95% CI 0.44-0.46; leave-one-study-out envelope 0.44-0.49). The depletion is absent among fully tryptic peptides (OR 1.06) and concentrates in missed-cleavage peptides (OR 0.62-0.57), is invariant to detected-abundance stratification, and retains its direction under three independent pI-calculator models. Case studies reveal both directions of precursor-centric masking. Two orthogonal validations ground the framework: a digestion-free synthetic-peptide analysis (basic-class odds ratio 0.73, 95% CI 0.71-0.76 under standard DDA) and a paired top-down benchmark, in which detected regions recover top-down sequence space (90% of spans overlapped) but do not delimit proteoform boundaries (8% containment). Pipeline, data, and code are public.