Module 4: Proteomics Identification & Quantification: Database Searching, Target-Decoy FDR & LFQ

1. Peptide-Spectrum Matching (PSM) & Database Searching

In shotgun discovery proteomics, acquired tandem mass spectra (MS2) are matched against theoretical in silico digests of the UniProt human proteome (Canonical + Isoforms). Algorithms like SEQUEST, Andromeda, Comet, and MSFragger compute cross-correlation scores ($\text{XCorr}$, hyperscore) between observed experimental fragment masses ($b$- and $y$-ions) and calculated fragmentation patterns:

• Precursor & Fragment Tolerance: High-resolution Orbitrap instruments utilize narrow precursor tolerance ($\pm 10\text{ ppm}$) and fragment tolerance ($\pm 0.02\text{ Da}$), drastically reducing the search space and random matches.
• Fixed vs. Variable Modifications: Carbamidomethylation of cysteine ($+57.0215\text{ Da}$) is treated as fixed (wet-lab alkylation with iodoacetamide). Variable modifications include methionine oxidation ($+15.9949\text{ Da}$), N-terminal protein acetylation ($+42.0106\text{ Da}$), and phospho-serine/threonine/tyrosine ($+79.9663\text{ Da}$).
• Enzyme Cleavage Specificity: Trypsin cleaves specifically at the C-terminus of lysine (K) and arginine (R), unless followed by proline (P). Allowing up to 2 missed cleavages accounts for incomplete enzymatic digestion.

2. The Target-Decoy Approach for False Discovery Rate (FDR) Control

To mathematically guarantee confidence in identified peptides, search engines append an artificial "decoy" database—created by reversing or shuffling the real protein sequences—to the real "target" database:

$$\text{FDR} = \frac{2 \times N_{\text{decoy}}}{N_{\text{target}} + N_{\text{decoy}}}$$

By adjusting score thresholds to maintain a strict $1\%$ peptide-spectrum match (PSM) FDR and a $1\%$ protein-level FDR, we suppress false-positive identifications across deep cancer proteomic callsets.

3. Label-Free Quantification (LFQ) & Protein Inference

Quantifying proteins without isotope labeling (LFQ) measures the extracted ion chromatogram (XIC) peak area of precursor ions across retention time. Advanced algorithms like MaxLFQ compute pairwise peptide ratio distributions to reconstruct robust protein intensities despite missing peptide measurements across samples:

# Automated discovery proteomics search with MSFragger and Philosopher CLI
philosopher workspace --init
philosopher database --custom ./uniprot_human_2026.fasta --contam

# Execute fast closed search on patient cohorts
msfragger ./closed_search.params ./converted_mzml/*.mzML

# Filter at 1% PSM and 1% Protein FDR with PeptideProphet / ProteinProphet
philosopher peptideprophet --database ./uniprot_human_2026.fasta --ppm --accmass *.pepXML
philosopher proteinprophet interact.pep.xml
philosopher filter --psm 0.01 --protein 0.01

# Generate LFQ intensity quantification report
philosopher freequant --dir ./converted_mzml/