CIForm

Introduction

CIForm, a Transformer-based model, can annotate cell types.

Instructions and examples are provided in the following tutorials.

Requirement

Python 3.9.12
PyTorch >= 1.5.0
numpy
pandas
scipy
sklearn
Scanpy
random

Input file

reference dataset.

cell type label of reference dataset.

query dataset.

Output file

After training the CIForm model, the model will be save at: "log/CIForm.tar".
The model prediction is saved in the log/y_predicts.npy.

Usage

import CIForm as CI
pred_result = CI.ciForm(s, referece_datapaths, Train_names, Testdata_path,Testdata_name)

in which

• s=The length of sub-vector,
• referece_datapaths=The path of annotated scRNA-seq datasets
• Train_names=The name of annotated scRNA-seq datasets
• Testdata_path=The path of query scRNA-seq datasets
• Testdata_name=The name of query scRNA-seq datasets

It is recommended that The label file be in the same directory as the corresponding data set and be named Labels.csv The label file should be a n rows * 1 column vector. For example,

Tutorial

• Cell Type annotation on Intra-datasets
• Cell Type annotation on Inter-datasets
• Cell Type annotation on Inter-datasets using multi-source

processsed Data

• Pancreas datasets(Baron Mouse,Baron Human,Xin,Muraro,Segerstolpe), TM(Tabula Muris), Zhang 68K, AMB .

• https://doi.org/10.5281/zenodo.3357167.

• Immune datasets(Oetjen, Dahlin, pbmc_10k_v3, and Sun[52]) and Brain datasets(Rosenberg, Zeisel, Saunders).

• https://figshare.com/articles/dataset/Benchmarking_atlas-level_data_integration_in_single-cell_genomics__integration_task_datasets_Immune_and_pancreas_/12420968.

The scRNA-seq datasets pre-processing code

• https://github.com/zhanglab-wbgcas/CIForm/tree/main/Data_processing

Paper Link

CIForm as a Transformer-based model for cell-type annotation of large-scale single-cell RNA-seq data https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbad195/7169137