pylimma documentation

pylimma is a faithful Python port of R limma, the most widely used Bioconductor package for differential expression analysis. It provides the full linear-modelling pipeline (lm_fit, contrasts_fit, e_bayes, top_table), voom for RNA-seq, gene-set testing (camera, roast, fry, romer), normalisation, batch correction, and differential splicing - all validated to match R limma output within rtol=1e-6 on fixture parity tests.

pylimma accepts numpy arrays, pandas DataFrames, AnnData objects, or limma-style EList dict subclasses, with centralised polymorphic dispatch so the same code works across the scverse ecosystem and R-style workflows.

Quickstart

import numpy as np
from pylimma import lm_fit, contrasts_fit, e_bayes, top_table

expr   = np.random.normal(size=(100, 6))
design = np.column_stack([np.ones(6), [0, 0, 0, 1, 1, 1]])

fit = lm_fit(expr, design)
fit = contrasts_fit(fit, contrasts=np.array([[0], [1]]))
fit = e_bayes(fit)
print(top_table(fit, coef=0).head())

See Quickstart for the full walk-through.

Contents

• Installation
  • Requirements
  • From PyPI
  • From source
  • R-parity testing
• Quickstart
  • Basic two-group comparison
  • AnnData workflow
  • EList workflow
  • Next steps
• API reference
  • Linear modelling
  • Voom and RNA-seq
  • Normalisation and batch
  • Duplicates, weights, correlation
  • Gene set testing
  • GO / KEGG enrichment
  • Statistical utilities
  • Model selection and mixture models
  • Splicing
  • Plotting
  • Data classes and dispatchers

Worked examples

• Differential Expression Analysis of the ALL Chiaretti Microarray Dataset with pylimma
  • Dataset
  • Pipeline
  • 1. Load the expression matrix + phenotype targets
  • 2. Per-sample distribution QC
  • 3. log-expression density (KDE)
  • 4. MDS (coloured by BT)
  • 5. Design matrix
  • 6. Fit linear models and contrasts (no voom)
  • 7. Empirical Bayes moderation
  • 8. Top table
  • 9. MD + volcano plots
  • 10. Heatmap of top 50 DE probes
• Differential Expression Analysis of Mouse Mammary RNA-seq with pylimma
  • Dataset
  • Pipeline
  • 1. Load counts and sample metadata
  • 2. Library sizes
  • 3. Filter low-expression genes
  • 4. log-CPM density before / after filter
  • 5. MDS (coloured by group)
  • 6. Design matrix
  • 7. voom
  • 8. Fit linear models and contrasts
  • 9. Empirical Bayes moderation
  • 10. Top table
  • 11. MD plot
  • 12. Volcano plot
  • 13. Heatmap of top 50 DE genes
• Differential Splicing Analysis with pylimma (pasilla dataset)
  • Dataset
  • Pipeline
  • 1. Load exon counts and sample metadata
  • 2. Library sizes
  • 3. log-count density (KDE)
  • 4. log2 transform
  • 5. Design matrix
  • 6. Fit per-exon linear models
  • 7. e_bayes + diff_splice + top_splice
  • 8. Top-ranked gene’s exon-level fold changes
• Differential Expression Analysis of Yoruba HapMap RNA-seq with pylimma
  • Dataset
  • Pipeline
  • 1. Load counts and sample metadata
  • 2. Library sizes
  • 3. Filter low-expression genes
  • 4. log-CPM density before / after filter
  • 5. MDS (coloured by group)
  • 6. Design matrix
  • 7. voom
  • 8. Fit linear models and contrasts
  • 9. Empirical Bayes moderation
  • 10. Top table
  • 11. MD plot
  • 12. Volcano plot
  • 13. Heatmap of top 50 DE genes
• Single-cell pseudobulk differential expression with pylimma (Kang 2018 PBMC)
  • Pipeline
  • 1. Load the Kang 2018 AnnData
  • 2. tSNE visualisation
  • 3. Pseudobulk
  • 4. Pseudobulk DE with pylimma (per cell type)
  • 5. Top IFN-stimulated genes per cell type
  • References
• Differential abundance analysis of proteomics data with pylimma
  • Dataset
  • Pipeline
  • Loading the data
  • 1. Load the matrix and targets
  • 2. PCA QC
  • 3. Design matrix and contrast
  • 4. lm_fit -> contrasts_fit -> e_bayes
  • 5. Top differentially abundant proteins
  • 6. Volcano plot

Validation

• Parity with R limma
  • Worked R-vs-Python notebooks
  • Tolerances
  • Known differences from R limma
  • Publication figures
• Known differences from R limma
  • Accepted differences
• Fixture-first validation strategy
  • Why fixture-first
  • Where fixtures live
  • Tolerances
  • R and limma versions
  • Regenerating fixtures
  • Versioning fixtures alongside code
• Benchmarks
  • Dataset set
  • Pipelines
  • Reproducibility
  • Numerical parity
  • Runtime comparison

Citation

Please cite the original limma papers when using pylimma:

• Smyth, G. K. (2004). Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Statistical Applications in Genetics and Molecular Biology 3, Article 3.

• Ritchie, M. E., Phipson, B., Wu, D., Hu, Y., Law, C. W., Shi, W., and Smyth, G. K. (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research 43(7), e47.

• Law, C. W., Chen, Y., Shi, W., and Smyth, G. K. (2014). voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biology 15, R29.

A pylimma preprint / Zenodo DOI will be listed here once published. Until then, cite the GitHub repository and the version tag.

Indices

• Index

• Module Index

• Search Page