Physical and genetic interactions, and pathways databases were used for network creation with addition of up to 30 related genes. ChEMBL (https://www.ebi.ac.uk/chembl/) and DSigDB (http://dsigdb.tanlab.org/DSigDBv1.0/) databases were used for initial selection of drugs that interact with at least one target from gene/protein network. From ChEMBL only drugs with described mechanisms were selected. From DSigDB FDA approved drugs and kinase inhibitors databases were used to find drugs with tested inhibitory activity. DSigDB computational drug signatures databases were used to find drugs that inhibit, decrease phosphorylation, expression and stability, block interaction with ligand and promote degradation of target genes/proteins. To calculate anti-viral and anti-fibrotic scores for drugs we used L1000 binary drug-induced gene expression signatures in full space (http://amp.pharm.mssm.edu/L1000FWD/). Drug-induced signatures were compared with two disease specific gene expression signatures: differentially expressed genes (DEGs) in idiopathic pulmonary fibrosis (IPF) vs. healthy lung used for fibrosis score calculation, and DEGs in SARS-CoV-2 infected A549 and normal human bronchial epithelial (NHBE) cells used for viral score calculation. Vi-Fi scores were calculated according to this formula:
- — fibrotic or viral score for the drug;
- — score for individual gene from drug-induced signature: +1 if gene expression is upregulated in disease-specific signature, -1 if downregulated;
- — Boolean p-value: 1 if p-value for gene expression difference in disease-specific signature is <0.05, 0 otherwise;
- — number of upregulated (up) or downregulated (down) genes in drug-induced signature.
Negative score values represent anti-viral or anti-fibrotic predicted effect, while positive score represents pro-viral or pro-fibrotic.
All analysis was performed in Python and codes are available (Github).