Date
Attendees
- Francisco Fernandez (Deactivated)
- Bryan Jones (Deactivated)
- Abhinandan Raghavan (Deactivated)
- Kieran Todd (Deactivated)
Apologies
Discussion items
Time | Item | Who | Output |
|---|---|---|---|
| 17:00 - 18:00 (60 mins) | >Assay table: gaps and sanity check >Descriptors | >All | >Final version of assay table |
Minutes
Assay table & Descriptors
Link to assay table including changes: Assays
>As long as we have sequence information for a molecule, data generated from in silico predictions are not seen as critical information to be shared at this stage - these can be generated retrospectively. In this context, in silico prediction data is output information.
>If, sequence information cannot be shared both Computational descriptors (see list from Bryan) and Biophysical-Molecular MOE descriptors (see list from Kieran) can be used to some extent to describe the sequence and structure of a molecule.
>Since Biophysical-Molecular MOE descriptor information can also be used to predict Chemical and Structure stability risk, in the context of not being able to share sequence information this is input data and should be shared.
>The sense is that Biophysical-Molecular MOE descriptor information can be used to generate most Computational descriptors e.g. MOE can calculate DI based on Biophysical-Molecular MOE descriptors, so the focus should be on the inclusion of Biophysical-Molecular MOE descriptor information as data to be shared when sequence is not availabe.
>With regards standardisation, most Biophysical-Molecular MOE descriptor information can be generated from sequence (and structure) without needing MOE (CCG has made available how each descriptor is calculated). It will be important to define the comparability of descriptors using different software packages if relevant.
>We need a clear position on the feasibility of anonimising sequence descriptors - discussion to be picked up in the next two weeks once Abhi has had some internal discussions.
>Regarding numbers of molecules to be shared: many of the assay workflows at GSK and Lilly are relatively new with regards the molecules which have been patented. As such, assay data is likely to be missing for some of the highlighted assays. NB - this needs to be highlighted as a risk and a mitigation put in place.
Actions
All organisation representatives - provide approximate number of molecules for each Tier (irrespective of whether information on these is likely to be shared).
Abhi - provide sequence descriptor examples and how PCA could be used to 'abstract' sequences.
All - contribute to assay table on wiki, internal stakeholder document (MODIFY ONLINE) and presentation (DOWNLOAD CURRENT VERSION AND MODIFY - SEND NEW VERSION TO RICHARD)