bvnGPS: The official implementation of "bvnGPS: a generalizable diagnostic model for acute bacterial and viral infection using integrative host transcriptomics and pretrained neural networks".
Input:
Host gene expression cohorts from Gene Expression Omnibus (The data in the paper could be downloaded and labeled with load_data/download_and_save_data.R)
Output:
- bacterial GPS, viral GPS, and noninfected GPS;
- decision tree, random forest, and support vector machines based on GPS selected;
- AUC performance of decision tree, random forest, and support vector machines on internal test data and external validation data. (Other performance could be calculated with
visualization/, such as accuracy, sensitivity, specificity, and ROC curve)