Rapid Focus in Lead Optimisation
We were asked to help a client with a difficult project to identify an orally bioavailable therapy for a CNS target. To date, they had not identified an active compound that had an acceptable combination of both oral bioavailability and CNS penetration. The client provided all project data and wanted to know how StarDrop could have helped them to identify suitable compounds at an earlier stage and how much resource its application would have saved.
Original Project Progress
In consultation with the client, an analysis of existing project data was performed to establish the major decision criteria for compound progression. The chronological progress of the project towards its target project profile was then charted.
The first 200 compounds progressed to in vitro ADME profiling (blue) were predominantly in one discreet area of chemical space. This was where the most potent compounds were located. However, as can be seen from the two examples, compounds in this area typically possessed either good bioavailability or good CNS penetration, but not both. In striving for greater potency the project chemists had isolated themselves in an area of chemistry that was unlikely to yield successful compounds.
A scoring profile was created in StarDrop to identify those compounds having the best overall balance of ADME properties based on in silico predictions. In this chemical space plot, high scoring compounds are coloured in light yellow and those with a poor predicted balance of ADME properties are coloured in red. Had this in silico analysis been carried out prior to compound synthesis, it would have highlighted the difficulties in this area of chemical space, potentially saving “misdirected” resource.
The second 200 compounds progressed to in vitro ADME profiling (green) showed a marked change in synthetic focus. Compounds in this lower region of the chemical space maintained good levels of activity but had a much better balance between bioavailability and CNS penetration as illustrated by the two compounds highlighted. However, to get to this point the client had synthesised and screened in excess of 3000 compounds, measured in vitro ADME for 400 compounds and progressed 70 compounds to in vivo pharmacokinetic analysis.
A retrospective analysis of the project shows how 25 compounds could have been identified for in vivo pharmacokinetic analysis from an initial virtual library of 3,100 molecules.
An initial subset of 300 compounds would be selected from the virtual library. Although the library had been designed with target activity in mind, little was known about the potency SAR and initial compound selection needed to cover a wide chemical space in order to identify diverse hits. However, by using StarDrop it was possible to profile the entire library for predicted ADME properties and include in the selection a bias towards compounds likely to have the required balance of ADME properties.The selected compounds would then be synthesised and screened for in vitro potency. Compounds were then scored again; this time for an appropriate balance between good potency (measured) and good ADME (predicted). A further subset of 25 compounds was then selected for progression in vivo. Here the bias in the selection was in favour of compounds having the best overall probability of success but with some degree of diversity included to aid “back-up” or “second series” identification. Whilst in vivo PK data are not available on all of the StarDrop selected compounds, it can be seen in this last plot that the approach selected key compounds that summarised the progress of the project in relation to its ADME target profile and also highlighted an area of space previously overlooked by the project team.