scMinerva: an Unsupervised Graph Learning Framework with Label-efficient Fine-tuning for Single-cell Multi-omics Integrated Analysis

Repository for manuscript scMinerva

Authors: Tingyang Yu, Yongshuo Zong, Yixuan Wang, Xuesong Wang, Yu Li

scMinerva is an unsupervised framework for single-cell multi-omics integrated analysis. The learned embeddings from the multi-omics data enable accurate integrated classification of cell types and stages. The power of scMinerva is sparkled by easy fine-tuning and is not sensitive to the using label size for fine-tuning a separate classifier. Out method could achieve excellent performance even with only 5% labeled data.

To use scMinerva, do the following:

Install the environment
Prepare the data
Train and evaluate scMinerva

Install the Environment

We provide a yml file containing the necessary packages for scMinerva. Once you have conda installed, you can create an environment as follows:

conda env create --file scMinerva.yml

Prepare the data

Required Files:

Omics: .csv file with shape (a,b) where a is the number of sample and b is the number of feature.
label: labels should be indexed start from 0 and be consecutive natural integers.
name the omics files: Files for omics features are supposed to be named as "i.csv" where i is an integer to distinguish omics. i.e. "2.csv".
name the label files: Label file is supposed to be named as "labels.csv" under the corresponding dataset directory.\

Train and evaluate scMinerva

Demo command on datase GSE156478_CITE:

python main.py --data_folder 'GSE156478_CITE' --num_omics 2 --num_class 7 --labeled_ratio 0.05

Parameters:

data_folder: the folder that contains prepared omics data and label data.
num_omics: the amount of omics inputed.
num_class: The number of class to classify.
labeled_ratio: Float number from 0 to 1, it means the proportion of labeled data for fine-tuning.

How to Cite

@article{yu2022scminerva,
  title={scMinerva: an Unsupervised Graph Learning Framework with Label-efficient Fine-tuning for Single-cell Multi-omics Integrated Analysis},
  author={Yu, Tingyang and Zong, Yongshuo and Wang, Yixuan and Wang, Xuesong and Li, Yu},
  journal={bioRxiv},
  pages={2022--05},
  year={2022},
  publisher={Cold Spring Harbor Laboratory}
}

Contact Us

Please open an issue or contact tyyistyu@gmail.com with any questions.# scMinerva

YistYU / scMinerva