BIOTIC (A Bayesian model to Integrate single-cell Multi-Omics for Transcription Factor Activity Inference to Characterize Cell Identity) is an innovative Bayesian model for multi-omics data that incorporates priori knowledge of the gene regulation process into generation process, and the causal relationship of gene regulation into the inference process.
https://github.com/Ying-Lab/BIOTIC
BIOTIC is generated by Python 3.8 and the R 4.2.1 is used during preprocessing
1.Install pytorch according to your computational platform
2.Install BIOTIC
git clone https://github.com/Ying-Lab/BIOTIC
cd BIOTIC
pip install -r requirements.txt refer to demo in a549 demo.ipynb
input: the count matrices(numpy) of gene counts, gene accessibility score, cell label, TF-target gene reference
output: trained model
other parameters:
learning_rate : learning rate for Adam optimizer, default 1e-3.
decay_rate : decay rate for Adam optimizer, default 0.97.
beta_1 : beta-1 parameter for Adam optimizer, default 0.99.
cuda : use GPU(s) to speed up training, default True.
float64 : use double float precision, default False.
config_enum : parallel, sequential or none. uses parallel enumeration by defaultdefault "parallel".
num_epochs : number of epochs to run, default 50.
decay_epochs : decay learning rate every #epochs, default 20.
batch_size : number of images (and labels) to be considered in a batch ,default 60.