This work gives us confidence in combining ADME models to calculate early human dose predictions and is being utilized in J&J discovery projects to help advance projects and facilitate decision-making.

Using a clever approach, Dries and the team demonstrated that the models we evaluated and propagated error allowed for ranking of compounds with 10-fold resolution. In the early drug discovery stage, this is very useful!

Thus, we focus on a method called 'early human dose prediction,' which uses ADME machine learning model parameters as input. This work addresses a key problem: If each ADME model has error, how does that impact the confidence around early human dose prediction?

You should make a chem / med chem starter pack @krhornberger.bsky.social