single-cell-proteomics-peptide-identification

Peptide and Protein Identification

Search MS2 spectra against a protein sequence database to identify peptides and proteins in your sample. Apply target-decoy FDR filtering to control false discovery rate at both PSM and protein levels.

This is Step 2 of the proteomics pipeline — takes centroided mzML from Step 1, produces PSM tables and protein groups for Step 3 (quantification).

---

What it does

1. Prepares the protein database: appends decoy sequences (reverse or scrambled) and common contaminants
2. Configures search parameters: enzyme specificity, variable and fixed modifications, mass tolerances
3. Runs database search with MSFragger (recommended) or Comet
4. Applies PSM-level FDR filtering using Percolator rescoring or classical target-decoy approach
5. Performs protein inference with parsimony principle to resolve shared peptides
6. Filters protein groups to 1% FDR
7. Exports results as TSV tables, pepXML, and mzIdentML
8. Generates summary statistics: number of PSMs, unique peptides, and protein groups

---

Why this exists

If you ask a general AI to "identify peptides in my MS data," it will:

• Not explain the target-decoy strategy or why it is required for FDR estimation
• Use incorrect MSFragger command-line flags (the CLI changed between v3 and v4)
• Skip protein inference entirely, leaving only peptide-level results
• Not distinguish between PSM FDR, peptide FDR, and protein FDR — applying only one threshold
• Not add contaminant sequences to the database, leading to misidentification of common lab proteins

This skill encodes the correct methodological decisions:

• Always appends a decoy database before searching (reversed sequences at minimum)
• Adds cRAP contaminant database (116 common laboratory contaminants)
• Distinguishes PSM FDR (1%) from protein FDR (1%) and applies both
• Uses parsimony protein grouping to handle shared peptides correctly
• Explains parameter choices for common modifications (oxidation M, carbamidomethyl C)