model icon

scgpt

Towards Building a Foundation Model for Single-Cell Multi-omics Using Generative AI

scGPT

This is the official codebase for scGPT: Towards Building a Foundation Model for Single-Cell Multi-omics Using Generative AI.

Preprint

Documentation

!UPDATE: We have released several new pretrained scGPT checkpoints. Please see the Pretrained scGPT checkpoints section for more details.

Installation

scGPT is available on PyPI. To install scGPT, run the following command:

$ pip install scgpt

[Optional] We recommend using wandb for logging and visualization.

$ pip install wandb

For developing, we are using the Poetry package manager. To install Poetry, follow the instructions here.

$ git clone this-repo-url
$ cd scGPT
$ poetry install

Note: The flash-attn dependency usually requires specific GPU and CUDA version. If you encounter any issues, please refer to the flash-attn repository for installation instructions. For now, May 2023, we recommend using CUDA 11.7 and flash-attn<1.0.5 due to various issues reported about installing new versions of flash-attn.

Pretrained scGPT Model Zoo

Here is the list of pretrained models. Please find the links for downloading the checkpoint folders. We recommend using the whole-human model for most applications by default. If your fine-tuning dataset shares similar cell type context with the training data of the organ-specific models, these models can usually demonstrate competitive performance as well.

Model name Description Download
whole-human (recommended) Pretrained on 33 million normal human cells. link
brain Pretrained on 13.2 million brain cells. link
blood Pretrained on 10.3 million blood and bone marrow cells. link
heart Pretrained on 1.8 million heart cells link
lung Pretrained on 2.1 million lung cells link
kidney Pretrained on 814 thousand kidney cells link
pan-cancer Pretrained on 5.7 million cells of various cancer types link

Fine-tune scGPT for scRNA-seq integration

Please see our example code in examples/finetune_integration.py. By default, the script assumes the scGPT checkpoint folder stored in the examples/save directory.

To-do-list

  • Upload the pretrained model checkpoint
  • Publish to pypi
  • Provide the pretraining code with generative attention masking
  • Finetuning examples for multi-omics integration, cell type annotation, perturbation prediction, cell generation
  • Example code for Gene Regulatory Network analysis
  • Documentation website with readthedocs
  • Bump up to pytorch 2.0
  • New pretraining on larger datasets
  • Reference mapping example
  • Publish to huggingface model hub

Contributing

We greatly welcome contributions to scGPT. Please submit a pull request if you have any ideas or bug fixes. We also welcome any issues you encounter while using scGPT.

Acknowledgements

We sincerely thank the authors of following open-source projects:

Citing scGPT

@article{cui2023scGPT,
title={scGPT: Towards Building a Foundation Model for Single-Cell Multi-omics Using Generative AI},
author={Cui, Haotian and Wang, Chloe and Maan, Hassaan and Pang, Kuan and Luo, Fengning and Wang, Bo},
journal={bioRxiv},
year={2023},
publisher={Cold Spring Harbor Laboratory}
}

Related notebook

BioTuring

Monocle3 - An analysis toolkit for single-cell RNA-seq

Build single-cell trajectories with the software that introduced **pseudotime**. Find out about cell fate decisions and the genes regulated as they're made. Group and classify your cells based on gene expression. Identify new cell types and states and the genes that distinguish them. Find genes that vary between cell types and states, over trajectories, or in response to perturbations using statistically robust, flexible differential analysis. In development, disease, and throughout life, cells transition from one state to another. Monocle introduced the concept of **pseudotime**, which is a measure of how far a cell has moved through biological progress. Many researchers are using single-cell RNA-Seq to discover new cell types. Monocle 3 can help you purify them or characterize them further by identifying key marker genes that you can use in follow-up experiments such as immunofluorescence or flow sorting. **Single-cell trajectory analysis** shows how cells choose between one of several possible end states. The new reconstruction algorithms introduced in Monocle 3 can robustly reveal branching trajectories, along with the genes that cells use to navigate these decisions.

T-cells

NK-cells

More

BioTuring

Mixscape: Analyzing single-cell pooled CRISPR screens

Expanded CRISPR-compatible CITE-seq (ECCITE-seq) which is built upon pooled CRISPR screens, allows to simultaneously measure transcriptomes, surface protein levels, and single-guide RNA (sgRNA) sequences at single-cell resolution. The technique enables multimodal characterization of each perturbation and effect exploration. However, it also encounters heterogeneity and complexity which can cause substantial noise into downstream analyses. Mixscape (Papalexi, Efthymia, et al., 2021) is a computational framework proposed to substantially improve the signal-to-noise ratio in single-cell perturbation screens by identifying and removing confounding sources of variation. In this notebooks, we demonstrate Mixscape's features using pertpy - a Python package offering a range of tools for perturbation analysis. The original pipeline of Mixscape implemented in R can be found here.

Ionocytes

Hepatocytes

More

BioTuring

Identifying tumor cells at the single-cell level using machine learning - inferCNV

Tumors are complex tissues of cancerous cells surrounded by a heterogeneous cellular microenvironment with which they interact. Single-cell sequencing enables molecular characterization of single cells within the tumor. However, cell annotation—the assignment of cell type or cell state to each sequenced cell—is a challenge, especially identifying tumor cells within single-cell or spatial sequencing experiments. Here, we propose ikarus, a machine learning pipeline aimed at distinguishing tumor cells from normal cells at the single-cell level. We test ikarus on multiple single-cell datasets, showing that it achieves high sensitivity and specificity in multiple experimental contexts. **InferCNV** is a Bayesian method, which agglomerates the expression signal of genomically adjointed genes to ascertain whether there is a gain or loss of a certain larger genomic segment. We have used **inferCNV** to call copy number variations in all samples used in the manuscript.

Theca cells

B-cells

More