TCR-DeepInsight

Aims

We have previously developed huARdb currently in developing which collects single-cells immune profiling datasets including linked transcriptome and full-length TCR informations. However, one of the main obstacles in using single-cell immune profiling datasets for disease immunotherapy is the absence of a convenient reference atlas to access this information. Despite the growing potential of TCR engineering in this area, there is a significant challenge in identifying functional TCRs due to the lack of efficient computational tools to integrate the data. Nonetheless, these datasets offer a valuable resource to identify such TCRs and further advance TCR-T technology.

Detailed collection of datasets

We aims to build the most comprehensive atlas containing matched transcriptome and full-length V(D)J sequence of T and B cells in Cancer, Autoimmune diseases, and Infections.

• In 2023, we make the first release of an integrated datasets containing more than 1,000,000 hcT cells with full length TCR sequence.
• In 2024, we will release the second version of the atlas with more than 20,000,000 hcT cells with full length TCR sequence, which is available at Zenodo.

Introduction to TCR-DeepInsight

To robustly identify potential disease associated TCRα/β pairs considering both TCR sequence similarity and transcriptome features from million-level paired TCRα/β repertoire, we developed a deep-learning based framework named TCR-DeepInsight.

Documentation

The documentation for TCR-DeepInsight is available at ReadTheDocs.

Installation

Hardware requirement for TCR-DeepInsight includes

1. RAM: >16Gb for larger dataset
2. VRAM of CUDA-enabled GPU: >8Gb

Install by PyPI

pip install tcr-deep-insight
pip install faiss-cpu # for CPU version
pip install faiss-gpu # for GPU version

or you can install cpu or gpu version by

pip install "tcr-deep-insight[gpu]"
pip install "tcr-deep-insight[cpu]"

Install by source code

You can create a running environment using conda (Anaconda or Miniconda)

conda create -n tcr-deep-insight -f environment.yml
conda activate tcr-deep-insight
git clone git@github.com:WanluLiuLab/TCR-DeepInsight.git
cd TCR-DeepInsight

Install alternative pytorch version

Please see the PyTorch official website for installing GPU-enabled version of PyTorch.

# Testing if CUDA is available
import torch
print(torch.__version__)
print(torch.cuda.is_available())

TCR-DeepInsight was tested on PyTorch 1.13.1 and CUDA 11.7

Usage

In IPython, simply import the package to get started:

import tcr_deep_insight as tdi 

For detailed usage, Please see the Documentation for TCR-DeepInsight.

License

This project is licensed under the BSD 3-Clause License.