CASE
STUDIES
COVID Moonshot: PostEra’s machine learning platform in real world drug discovery


In March 2020 PostEra helped launch COVID Moonshot, which became the world’s largest open-science initiative to develop a COVID antiviral cure.
PostEra’s machine learning drove the prioritization of crowdsourced ideas
Moonshot crowdsourced antiviral ideas, inspired by a fragment screen, from over 400 scientists around the world. PostEra’s machine learning drove the prioritization of these ideas and designed the synthesis of all 2000 compounds made as part of COVID Moonshot as well as using our ML technology to develop new chemotypes.
Property | Target Range | Cold Start Mar 2020 --> Dec 2021 |
---|---|---|
Protease Assay | IC50 < 50 nM | 40nM |
Viral Replication | EC50 < 0.2µM | 0.15 μM in A549 CPE |
PK-PD | Cmin > EC90 (plaque reduction) for 24h | Current projected human dose ~220mg QD ; 100mg BID |
Coronavirus spectrum | SARS-CoV2 B1.1.7 , 501.V2, B.1.1.248 variants essential SARS-CoV-1 & MERS desirable | Active against B1.1.7 , 501.V2 in cellular assays |
Route of administration | oral | BO = 45% in rat |
Solubility | > 5 mg/mL, >100µM tolerable | 750 µM |
Half-Life | Ideally >= 8 h (human) est from rat and dog | Rat 2h, human predicted PK sufficient |
Safety | No significant protease activity >50% at 10μM (Nanosyn 61 protease panel) Only reversible and monitorable toxicities (NOAEL > 30x Cmax) No significant DDI – clean in 5 CYP450 isoforms Critical transporter check (e.g. OATP) hERG and NaV1.5 IC50 > 50 µM No significant change in QTc No mutagenicity or teratogenicity risk | Protease panel clean on analogues Eurofins / CEREP 44 target panel clean Cyp450: clean except 2A4 (3uM) No hERG activity Live phase planned Lead compounds are clean in AMES +/- S9 |