Installation

Note

Hardware requirement for TCR-DeepInsight includes

  1. CPU: >= 1 cores. Recommended >= 8 cores for large-scale dataset

  2. RAM: >=1 Gb. Recommended >=64 Gb for large-scale dataset

  3. VRAM >= 1Gb of CUDA-enabled GPU. Recommended >= 8 Gb for large-scale dataset

  4. Disk space >= 1Gb. Recommended >= 100Gb for large-scale dataset

Operation System requirements for running TCR-DeepInsight include the installation of Python3 (Python3.9 used for development) and several PyPI packages. You can create a running environment using conda.

Install from PyPI

1  pip install tcr-deep-insight

Install from source

1  conda create -n tcr-deep-insight -f environment.yml
2  conda activate tcr-deep-insight
3  git clone git@github.com:WanluLiuLab/TCR-DeepInsight.git
4  python3 setup.py install

Usage

In IPython, simply import the package to get started:

1  import tcr_deep_insight as tdi

For more details, please refer to the Tutorial.

Example datasets

Warning

TCR-DeepInsight require AnnData objects as input.

Example processed datasets are available at Zenodo.

or you can use the following code to download the example dataset:

1import tcr_deep_insight as tdi
2gex_adata = tdi.data.human_gex_reference_v2()
3tcr_adata = tdi.data.human_tcr_reference_v2()

Pretrained models

Pretrained models are available at Zenodo.

or you can use the following code to download the pretrained model:

1import tcr_deep_insight as tdi
2tdi.data.download_model_weights()