Identification of acrylamide-based covalent inhibitors of SARS-CoV-2 (SCoV-2) Nsp15 using high-throughput screening and machine learning.

Non-structural protein 15 (Nsp15) is a SARS-CoV-2 (SCoV-2) endoribonuclease and is a promising target for drug development because of its essential role in evading the host immune system. However, developing inhibitors against Nsp15 has been challenging due to its structural complexity and large RNA binding surface. In this report, we screened a 2640 acrylamide-based compound library against Nsp15 and identified 10 fragments that reacted with cysteine residues on Nsp15 and inhibited its endoribonuclease activity with IC50s less than 5 μM. These compounds had several attractive properties, such as low molecular weight (180-300 g mol-1), log P <3, zero violations to Lipinski's rules, and no apparent pan-assay interference (PAINs) properties. In addition, based on this data as a training set, we developed an artificial intelligence (AI) model that accelerated the hit to lead process and had a 73% accuracy for predicting new acrylamide-based Nsp15 inhibitors. Collectively, these results demonstrate that acrylamide fragments have great potential for developing Nsp15 inhibitors.