Higashi Requirements

Installation

There are two ways of using Higashi:

1. clone the repo and run Higashi program by running python ....py (referred to as Higashi(CLI)).
2. use higashi as a package which allows you to include Higashi in your custom scripts or use it with jupyter notebook by wrapping all Higashi functions in a python class (referred to as Higashi(API)).

Method 1: for Higashi(CLI) & Higashi(API)

1. Install all the dependencies listed below
2. git clone https://github.com/ma-compbio/Higashi/
3. Run Higashi.
4. If you want to use Higashi in API style, put the higashi directory of this repo under your project directory

Method 2: for Higashi(API)

1. Install pytorch with CUDA support when applicable
2. conda install -c ruochiz higashi

If you run into any error when installing Higashi with conda, please describe your OS system, conda version, python, pytorch version when submitting the issue.

Requirements

Higashi

• Python (>=3.5.0, tested on 3.7.9)
• h5py (tested on 2.10.0)
• numpy (tested on 1.19.2)
• pandas (tested on 1.1.3)
• pytorch (>= 1.4.0)
• fbpca (tested on 1.0.0)
• scikit-learn (tested on 0.23.2)
• tqdm (tested on 4.50.2)

Output to cool/scool

• Cooler (tested on 0.8.11)

Generating visualization plots (Higashi-analysis/Higashi-vis)

• seaborn
• matplotlib
• umap-learn

Interactive visualization sessions (Higashi-vis)

• bokeh (tested on 2.1.1)
• PIL (tested on 7.2.0)
• cachetools (tested on 4.1.1) (Optional if you select the same cell in a short period of time, cachetools would display the cached image instead of rendering a new one. )
• cmocean (Optional for beautiful cmaps.)

Quick start:

• Input Files
• Higashi Usage
• Output of Higashi main