Computes phenotypes using SiCNMF algorithm described in
Please cite above work if using this code.
REQUIREMENTS: python, python packages: numpy, scipy, sklearn
HOW TO USE:
- Fill in the input parameters in SiCNMF.config file (should be in the same directory as SiCNMF_start.py). Replace the sample inputs in the file.
- Run the following command in a terminal: python SiCNMF_start.py
INPUTS: ---model: CNMF for no sparsity inducing factors or SiCNMF ---sources: Comma separated list of sources of patient data to be used in SiCNMF ---Xfiles: Comma separated list of flat file names containing EHR data. Each file should be an .npy file containing a sparse matrix of EHR data of patients x source data. Check: numpy.load(<.npy filename>).ravel()[0] should return a sparse matrix of size nPatients x nSourceItems. ---nFactors: Number of latent factors/phenotypes to return ---[optional]outdir: Relative path to the directory to save output files. Default='./' ---[optional]etaSweep: Comma separated list of paramter values of eta to evaluate SiCNMF on. Default=[1]
OUTPUTS: