tooluniverse-proteomics-analysis

What it does

Analyzes mass spectrometry proteomics data to quantify proteins, identify differential expression between conditions, analyze post-translational modifications (PTMs), and study protein-protein interactions. It handles outputs from multiple MS platforms (MaxQuant, Spectronaut, DIA-NN, and others), performs normalization and statistical analysis, conducts pathway enrichment, and can integrate proteomics with transcriptomics.

How it works

The skill provides a workflow that starts with importing MS proteomics data (e.g., MaxQuant proteinGroups.txt, Spectronaut *_Report.tsv, DIA-NN report.tsv), builds an intensity matrix, and attaches protein metadata. It includes quality control steps such as missing value assessment, intensity distribution checks, sample correlation, and PCA. Preprocessing filters out low-confidence proteins, imputes missing values (MinProb, KNN, QRILC options), and normalizes across samples (median, quantile, TMM, or VSN methods).

Differential expression is performed using a limma-like approach: computing log2 fold changes between two groups, t-tests with unequal variance, and multiple testing correction (BH). Results yield log2FC, p-values, adjusted p-values, and a significance flag for |log2FC| > 1 and adj_p_value < 0.05. A volcano plot visualization is provided.

Phase 4 covers PTM analysis, including peptide-level modification identification, localization filtering, site quantification, and kinase-substrate prediction. Phase 5 performs functional enrichment (GO, KEGG/Reactome, CORUM) and Phase 6 analyzes protein-protein interactions via networks and clustering. Phase 7 supports optional multi-omics integration (e.g., correlation with RNA-seq) followed by a Phase 8 report generation with summary statistics and visualizations.

The skill includes concrete code snippets for data loading, QC plots, filtering logic, missing-value imputation, normalization, differential expression computation, and a volcano plot, illustrating the concrete steps the agent should execute when given appropriate inputs.

When to use it

Use when you have proteomics MS data and you want to compare protein abundance between conditions, analyze PTMs, study protein complexes or interactions, integrate with RNA data, or discover proteomic biomarkers. Triggers include requests for differential protein expression, PTM analysis, protein-RNA correlations, multi-omics integration, PPI analysis, and biomarker discovery.

What it can touch

The skill specifies inputs and outputs for importing MS data formats (MaxQuant, Spectronaut, DIA-NN, Proteome Discoverer). It references functions that load data, perform QC, normalization, differential testing, PTM analyses, kinase predictions, enrichment analyses, and network visualizations. It uses tool names like "MaxQuant proteinGroups.txt", "Phospho (STY)Sites.txt", and programmatic blocks including Python code for data loading, QC, and analysis. It also mentions interacting with a tool ecosystem for kinase prediction (e.g., hypothetical functions) and STRING/IntAct for PPI networks.

Caveats

The skill declares a NOASSERTION license. It outlines standard limitations of proteomics analyses (e.g., missing values, choice of imputation and normalization methods, multiple testing corrections, and potential biases in PTM localization). It provides example thresholds (p < 0.05, |log2FC| > 1) and notes conditional phases (e.g., PTM analysis is optional if applicable).