Plasma Proteomics Differential Analysis & GSEA Pipeline

Overview

This repository contains a reproducible workflow for analyzing plasma proteomics data using:

• Linear modeling (limma) for differential protein expression
• Gene Set Enrichment Analysis (GSEA, preranked via gseapy)

The pipeline is designed for multi-cohort clinical proteomics data with repeated measures and confounding variables.

---

🧬 Data Description

1. sample_info.tsv

Metadata table describing each sample.

Column	Description
Sample ID	Patient identifier (prefix indicates country: SWE, IRE, ITA, NOR)
TP	Timepoint (T0, T1, T2)
Sex	Biological sex (M / F)
CF ID	Unique sample identifier (used to map quantification data)
cancer_type	Cancer type
toxicity	Toxicity status (yes / no)
country	Country of origin
...	Additional clinical covariates

---

2. meta.tsv

Mapping table linking CF IDs to raw file names.

Column	Description
Sample ID	CF ID
R.FileName	Identifier used in quantification matrix

---

3. plasma_prot_quant.tsv

Protein quantification matrix.

Column	Description
PG.Genes	Protein / gene annotation
Other columns	Sample-specific intensities (mapped via meta.tsv)

---

⚙️ Data Processing Workflow

1. Column Mapping

• Match quantification columns to CF IDs using meta.tsv
• Rename columns to CF IDs

2. Gene Name Cleaning

• Multi-annotation entries (e.g. ARHGEF5;ARHGEF5;...) are simplified to:

3. Expression Matrix Construction

• Rows → samples (CF IDs)
• Columns → proteins
• Values → normalized intensities

4. Preprocessing

• Remove duplicated samples
• Convert to numeric
• Log2 transform

---

📊 Differential Expression Analysis (limma)

We use the limma framework for linear modeling with empirical Bayes moderation.

Key Features

• Supports flexible contrasts:
• Sex: M vs F
• TP: T2 vs T0
• toxicity: yes vs no
• Adjusts for confounders:
• country
• cancer_type
• TP
• toxicity

---

Example Usage

results <- limma_differential(
expr = expr_matrix,
sample_info = sample_info,
contrast_feature = "Sex",
contrast_direction = "M_vs_F",
confounding_factors = c("country", "cancer_type", "TP", "toxicity")
)