proteomics-quantification

proteomics-quantification

When to use

The user has a peptide / PSM table and wants protein-level abundance via one of:

• lfq (default) — Label-Free Quantification by intensity summation. Requires an intensity column.
• ibaq — intensity-Based Absolute Quantification (intensity / theoretical tryptic peptide count). Requires an intensity column AND ONE OF: a per-protein sequence column (in-silico digested by the script) OR a pre-computed n_theoretical_peptides integer column. Without either, the script silently estimates unique_peptides × 1.5.
• spectral_count — PSM count per protein (no intensity needed).

Pick with --method {lfq,spectral_count,ibaq} (default lfq). For TMT / iTRAQ label-based workflows, perform the search-engine quant first; this skill is intensity- / count-only.

Inputs & Outputs

Inputs

• Modalities: lfq
• File types: .csv
• Accepts artifact proteomics.peptide_table (csv)

Outputs

• tables/protein_abundance.csv
• report.md
• result.json
• Produces artifact proteomics.abundance_matrix as tables/protein_abundance.csv (csv)

Flow

1. Load CSV (--input <peptides.csv>) or generate a demo (--demo).
2. Dispatch on --method (proteomics_quantification.py:156); validate required columns per method.
3. Aggregate per protein:
   • lfq: sum intensity per protein.
   • ibaq: sum intensity per protein, divide by n_theoretical_peptides. Source order at proteomics_quantification.py:115-130: sequence (compute on the fly) → n_theoretical_peptides (use as-is) → unique_peptides × 1.5 (silent estimate with warning).
   • spectral_count: count PSMs per protein.
4. Write tables/protein_abundance.csv (proteomics_quantification.py:277) + report.md + result.json (:283).

Gotchas

• lfq and ibaq require an intensity column; method enforces this. proteomics_quantification.py:77 raises ValueError("Input requires an 'intensity' column for LFQ"); :109 raises the same for iBAQ. spectral_count only needs row counts (no intensity).
• ibaq requires either sequence OR n_theoretical_peptides; otherwise it SILENTLY ESTIMATES. proteomics_quantification.py:115-130 checks for sequence first (in-silico digest at :42-72, K/R not before P, length 7-30), then n_theoretical_peptides, otherwise falls back to unique_peptides × 1.5 with only a logger warning. The wrong column name (theoretical_peptides instead of n_theoretical_peptides) silently triggers the estimate path — always pass one of the two correct columns.
• Unknown --method raises ValueError. proteomics_quantification.py:156 rejects values outside ("lfq", "spectral_count", "ibaq"). The argparse choices= already enforces this — the :156 raise is defence-in-depth for direct library calls.
• --input REQUIRED unless --demo. proteomics_quantification.py:269 raises ValueError("--input required").
• Missing intensities in lfq are summed as 0. pd.Series.sum(skipna=True) is the default — proteins with all-NaN intensities yield 0, indistinguishable from "all detected as zero". Pre-filter or impute upstream if NaN-vs-zero matters.