Recently, I was interested in an article of JMC.
The author analyzed in-house compound selection and found rule that Easy-to-understand scouring function AB-MPS.
AB-MPS is defined by following equation.
AB-MPS = Abs( cLogP – 3 ) + NAR + NRB
Where NAR means number of aromatic rings and NRB means number of rotatable bounds.
They found that AB-MPS of beyond the Ro5 compounds shows good correlation with Oral bioavailability (F) and some kinds of ADMET parameters.
It is not true everywhere but I think the parameter is good indicator for medicinal chemist because easy to understand and based on in-house dataset ( for author’s company ). We can make more complex predictive model by using machine learning method, but the method is difficult to understand why these compounds are good.
In house dataset is key factor of its strengths.
I am still thinking about how to collect in-house data and how to use these dataset more efficiently.