gwaslab 3.4.42

A collection of handy tools for GWAS SumStats

GWASLab

• A handy Python toolkit for handling GWAS summary statistics (sumstats).
• Each process is modularized and can be customized to your needs.
• Sumstats-specific manipulations are designed as methods of a Python object, gwaslab.Sumstats.

Please check GWASLab documentation at https://cloufield.github.io/gwaslab/

Note: GWASLab is being updated very frequently for now. I will release the first stable version soon! Please stay tuned.

Warning: Known issues of GWASLab are summarized in https://cloufield.github.io/gwaslab/KnownIssues/ .

Install

install via pip

pip install gwaslab==3.4.41

import gwaslab as gl
# load plink2 output
mysumstats = gl.Sumstats("t2d_bbj.txt.gz", fmt="plink2")

# load sumstats with auto mode (auto-detecting common headers) 
# assuming ALT/A1 is EA, and frq is EAF
mysumstats = gl.Sumstats("t2d_bbj.txt.gz", fmt="auto")

# or you can specify the columns:
mysumstats = gl.Sumstats("t2d_bbj.txt.gz",
             snpid="SNP",
             chrom="CHR",
             pos="POS",
             ea="ALT",
             nea="REF",
             neaf="Frq",
             beta="BETA",
             se="SE",
             p="P",
             direction="Dir",
             n="N",
             build="19")

# manhattan and qq plot
mysumstats.plot_mqq()
...

install in conda environment

Create a Python 3.9 environment and install gwaslab using pip:

conda env create -n gwaslab_test -c conda-forge python=3.9
conda activate gwaslab
pip install gwaslab==3.4.41

or create a new environment using yml file environment_3.4.40.yml

conda env create -n gwaslab -f environment_3.4.40.yml

install using docker

A docker file is available here for building local images.

Functions

Loading and Formatting

• Loading sumstats by simply specifying the software name or format name, or specifying each column name.
• Converting GWAS sumstats to specific formats:
  • LDSC / MAGMA / METAL / PLINK / SAIGE / REGENIE / MR-MEGA / GWAS-SSF / FUMA / GWAS-VCF / BED...
  • check available formats
• Optional filtering of variants in commonly used genomic regions: Hapmap3 SNPs / High-LD regions / MHC region

Standardization & Normalization

• Variant ID standardization
• CHR and POS notation standardization
• Variant POS and allele normalization
• Genome build : Inference and Liftover

Quality control, Value conversion & Filtering

• Statistics sanity check
• Extreme value removal
• Equivalent statistics conversion
  • BETA/SE , OR/OR_95L/OR_95U
  • P, Z, CHISQ, MLOG10P
• Customizable value filtering

Harmonization

• rsID assignment based on CHR, POS, and REF/ALT
• CHR POS assignment based on rsID using a reference text file
• Palindromic SNPs and indels strand inference using a reference VCF
• Check allele frequency discrepancy using a reference VCF
• Reference allele alignment using a reference genome sequence FASTA file

Visualization

• Mqq plot: Manhattan plot, QQ plot or MQQ plot (with a bunch of customizable features including auto-annotate nearest gene names)
• Miami plot: mirrored Manhattan plot
• Brisbane plot: GWAS hits density plot
• Regional plot: GWAS regional plot
• Genetic correlation heatmap: ldsc-rg genetic correlation matrix
• Scatter plot: variant effect size comparison
• Scatter plot: allele frequency comparison
• Scatter plot: trumpet plot (plot of MAF and effect size with power lines)

Visualization Examples

Other Utilities

• Read ldsc h2 or rg outputs directly as DataFrames (auto-parsing).
• Extract lead variants given a sliding window size.
• Extract novel loci given a list of known lead variants / or known loci obtained from GWAS Catalog.
• Logging: keep a complete record of manipulations applied to the sumstats.
• Sumstats summary: give you a quick overview of the sumstats.
• ...

Requirements (deprecated)

environment.yml

name: gwaslab
channels:
  - conda-forge
  - defaults
dependencies:
  - python=3.8.16=h7a1cb2a_3
  - jupyter==1.0.0
  - pip==23.1.2
  - pip:
      - adjusttext==0.8
      - biopython==1.81
      - gwaslab==3.4.16
      - liftover==1.1.16
      - matplotlib==3.7.1
      - numpy==1.24.2
      - pandas==1.4.4
      - scikit-allel==1.3.5
      - scikit-learn==1.2.2
      - scipy==1.10.1
      - seaborn==0.11.2
      - statsmodels==0.13
      - adjustText==0.8
      - pysam==0.19
      - pyensembl==2.2.3

How to cite

• GWASLab preprint: He, Y., Koido, M., Shimmori, Y., Kamatani, Y. (2023). GWASLab: a Python package for processing and visualizing GWAS summary statistics. Preprint at Jxiv, 2023-5. https://doi.org/10.51094/jxiv.370

Sample Data

• Sample GWAS data used in GWASLab is obtained from: http://jenger.riken.jp/ (Suzuki, Ken, et al. "Identification of 28 new susceptibility loci for type 2 diabetes in the Japanese population." Nature genetics 51.3 (2019): 379-386.).

Acknowledgement

Thanks to @sup3rgiu, @soumickmj and @gmauro for their contributions to the source codes.

Contacts

• Github: https://github.com/Cloufield/gwaslab
• Blog (in Chinese): https://gwaslab.com/
• Email: [e-mail]
• Stats: https://pypistats.org/packages/gwaslab