Anti-SARS-CoV-2 Ability Prediction
We construct machine learning models to quickly identify potential compounds actively targeting to COVID-19 virus.
|Viral Replication||3CL enzymatic activity|
|Viral Entry||ACE2 enzymatic activity|
|TMPRSS2 enzymatic activity|
|Spike-ACE2 protein-protein interaction AlphaLISA|
|Spike-ACE2 protein-protein interaction TruHit Counterscreen|
|Live Virus Infectivity||HEK293 cell line toxicity|
|Human fibroblast toxicity|
|SARS-CoV-2 cytopathic effect CPE|
|SARS-CoV-2 cytopathic effect host tox counterscreen|
Promising drug meet these criteria:
1. is active 3CL Protease inhibitor 2. is active ACE2 inhibitor 3. is not cytotoxic (HEK293 cell line and human fibroblast) 4. is active in CPE 5. is active in Spike/ACE2 6. is not active in the counterscreen 7. is active TMPRSS2 protease inhibitor
Predicition is implemented by machine learning modeling using k-nearest neightbors (KNN) algorithm.
Set the similarity threshold and see how many compounds could be found from our drug dataset.
Compounds that are > similar to input molecules will be selected from ZINC database.