English | 简体中文(推荐)

JanusX is a high-performance toolkit for Genome-Wide Association Studies (GWAS) and genomic selection, built on mixed linear models (MLM). It provides significant performance improvements over tools like GEMMA, GCTA, and rMVP, especially in multi-threaded computation.

git clone https://github.com/MaizeMan-JxFU/JanusX.git
cd JanusX
sh ./install.sh

The install script uses uv for dependency management and creates a virtual environment in .venv/.

For convenience, we also provide pre-compiled binaries that don't require building from source. The releases are available at Releases v1.0.0 for the following platforms:

• Linux AMD64: JanusX-v1.0.0-linux-AMD64.tgz
• Windows AMD64: JanusX-v1.0.0-windows-AMD64.tgz

Simply download and extract the archive, then run the executable directly.

Note: Windows installation is no longer supported. Please use Linux/macOS or Windows Subsystem for Linux (WSL). But there is a pre-build version for Windows.

./jx -h
./jx <module> [options]

Note that running ./jx -h might take a while at first! This is because the Python interpreter is compiling source code into the pycache directory. Subsequent runs will use the pre-compiled code and load much faster!

• gwas - Mixed linear model GWAS analysis
• postGWAS - Visualization and annotation
• transanno - Annotation genome version migration
• gformat - Genotype format conversion
• grm - Genetic relationship matrix calculation
• pca - Principal component analysis

# GWAS with VCF input
jx gwas --vcf example/mouse_hs1940.vcf.gz --pheno example/mouse_hs1940.pheno --out test

# GWAS with PLINK format
jx gwas --bfile genotypes --pheno phenotypes.txt --out results --grm 1 --qcov 3 --thread 8

# Using kinship matrix and fast mode
jx gwas --vcf genotypes.vcf --pheno phenotypes.txt --out results --grm kinship_matrix.txt --qcov 10 --lm

# Visualize GWAS results
jx postGWAS --files test/*.mlm.tsv --threshold 1e-6 --thread 4

# Genomic selection with PLINK format
jx gs --bfile genotypes --pheno phenotypes.txt --out results --GBLUP --rrBLUP --RF

Test data in example is from genetics-statistics/GEMMA, published in Parker et al, Nature Genetics, 2016

• pyBLUP - Core statistical engine

  • gwas.py - GWAS class implementing mixed linear model with REML optimization
  • QK.py - Q matrix (population structure) and K matrix (kinship) calculation with memory-optimized chunking (deprecated)
  • QK2.py - Alternative QK implementation
  • QC.py - Quality control functions (MAF, missing rate filters)
  • mlm.py - Mixed linear model utilities
  • pca.py - PCA computation (includes randomized SVD for large datasets)
  • kfold.py - Cross-validation utilities

• gfreader - Genotype file I/O

  • base.py - Readers for VCF, PLINK binary (.bed/.bim/.fam), HapMap, and numpy formats
  • Supports genotype conversion between formats

• bioplotkit - Visualization

  • manhanden.py - Manhattan and QQ plots
  • LDBlock.py - LD block visualization
  • gffplot.py - GFF/annotation plotting
  • pcshow.py - PCA visualization (uses Plotly)

Each module corresponds to a CLI command. The launcher script (jx.bat/jx) dispatches to module/<name>.py.

Mixed Linear Model: Uses eigen decomposition of the kinship matrix to simplify variance computation, with Brent's method for REML parameter optimization. Lambda (variance ratio) is the single parameter being optimized.

Kinship Methods: VanRanden (Centralization, default), Yang (Standardization).

PCA: Matrix block partitioning for computation.

Phenotype file: Tab-delimited, first column is sample ID, subsequent columns is phenotype

Supported genotype formats: VCF (.vcf, .vcf.gz), PLINK binary (.bed/.bim/.fam), numpy archives (.npz/.snp/.idv)

Requires Python 3.8+

Example data in example/ directory from Parker et al, Nature Genetics, 2016 (via GEMMA project)