Skip to content

GAISHI

gaishi is a Python package for Genomic Analysis of Introgressed-Site and -Haplotype Identification using machine learning. Currently, it supports three types of models:

  • Logistic regression models
  • Extra-trees classifiers
  • U-Net models

gaishi uses established population genetic simulators like msprime for generating training and test data. It can be applied to detect introgressed fragments or alleles in genomes from various species.

Requirements

gaishi works on UNIX/LINUX operating systems and tested with the following:

  • Python 3.9

  • Python packages:

    • demes=0.2.3
    • h5py=3.10.0
    • joblib=1.3.2
    • msprime=1.3.1
    • numpy=1.26.4
    • pandas=2.2.1
    • pydantic=2.11.7
    • python=3.9.19
    • pyranges=0.0.129
    • pytest=8.1.1
    • scikit-allel=1.3.7
    • scikit-learn=1.4.1.post1
    • scipy=1.12.0
    • tskit=0.5.6
    • pyyaml=6.0.1
    • seriate==1.1.2
    • torch==2.2.0

Installation

Users can install gaishi by using the following commands:

git clone https://github.com/xin-huang/gaishi
cd gaishi
mamba env create -f env.yaml
mamba activate gaishi
pip install .

Users first need to install mamba to create the virtual environment.

Help

To get help information, users can use:

gaishi -h

This will display information for three commands:

Command Description
lr Use logistic regression models
unet Use U-Net models
eval Evaluate model performance

If you need further help, such as reporting a bug or suggesting a feature, please open an issue.