limma-py 0.1.0

Python implementation of R's limma package for differential expression analysis

limma_py

A Python implementation of the popular R limma package for differential expression analysis of gene expression data.

Overview

limma_py is a comprehensive Python port of the widely-used R limma (Linear Models for Microarray Data) package. It provides powerful statistical methods for analyzing gene expression data from microarray and RNA-seq experiments, with a focus on differential expression analysis.

Key Features

• ✅ Complete Functionality - Provides the same core functionality as R's limma package
• ✅ Linear Model Fitting - Supports complex experimental design matrices
• ✅ Empirical Bayes Moderation - Improves statistical power for small sample data
• ✅ Multiple Group Comparisons - Supports complex contrast analysis
• ✅ Pythonic API - User-friendly interface designed for Python users
• ✅ High Performance Computing - Optimized implementation based on NumPy and SciPy

Installation

Via pip

pip install limma_py

From source

git clone https://github.com/wd1566/limma_py
cd limma_py
pip install -e .

Quick Start

import pandas as pd
import numpy as np
import limma_py

# Read CSV file, header=0 indicates the first row contains column names
data = pd.read_csv("data/Harmine_iTSA.csv", header=0)

# Extract gene names (first column)
gene_names = data.iloc[:, 0].values

# Extract expression matrix (all columns starting from the second)
expr_data = data.iloc[:, 1:]

# Define experimental groups: first 5 samples are control group (V_group), last 5 are treatment group (D_group)
group = np.array(["V_group"] * 5 + ["D_group"] * 5)

# Create design matrix using one-hot encoding
design_df = pd.get_dummies(group, drop_first=False)[["V_group", "D_group"]]

# Ensure design matrix is of integer type
design = design_df.astype(int)

# Create a copy of the expression matrix and set gene names as row indices
expr_matrix = expr_data.copy()
expr_matrix.index = gene_names

# Perform linear model fitting using limma
fit_python = limma_py.lmFit(expr_matrix, design)

# Set up contrast matrix: compare differences between treatment group (D_group) and control group (V_group)
contrasts = limma_py.make_contrasts('D_group - V_group', levels=design)

# Perform contrast analysis on the fitted results
fit_python = limma_py.contrasts_fit(fit_python, contrasts)

# Moderate standard errors using empirical Bayes method
eb_python = limma_py.eBayes(fit_python)

# Extract differential expression analysis result table
res = limma_py.toptable(eb_python)

Data Format Requirements

Input data should be in CSV format with the following structure:

Gene,Sample1,Sample2,Sample3,Sample4,Sample5,Sample6,Sample7,Sample8,Sample9,Sample10
Gene_1,8.45,7.89,8.12,15.67,14.89,16.23,8.33,7.95,8.21,8.09
Gene_2,12.34,11.89,12.56,11.45,10.98,12.11,25.67,24.89,26.01,25.34
Gene_3,5.67,6.01,5.89,18.90,19.45,18.23,6.12,5.78,6.34,5.95

Format specifications:

• First column: Gene/protein names (any identifier)

• Subsequent columns: Numerical expression matrix, each column represents a sample

• Must be in CSV format, other formats need to be converted first

Core Functions

• lmFit() - Linear model fitting for gene expression data
• make_contrasts() - Generate contrast matrices for group comparisons
• contrasts_fit() - Extract results for specified contrasts
• eBayes() - Empirical Bayes moderation of standard errors
• toptable() - Extract top-ranked genes from analysis results

Documentation

For detailed documentation, examples, and API reference, visit our documentation site.

• Installation Guide
• Usage Examples
• API Reference

Use Cases

• Microarray Data Analysis
• RNA-seq Differential Expression Analysis
• Multiple Group Experimental Designs
• Time Series Expression Data
• Any expression data analysis requiring linear modeling

Dependencies

• pandas >= 1.0.0
• numpy >= 1.18.0
• scipy >= 1.4.0
• statsmodels >= 0.11.0

Contributing

Contributions are welcome! Please feel free to submit pull requests, report bugs, or suggest new features.

License

This project is licensed under the MIT License - see the LICENSE file for details.