2018-05-01 Project Team Meeting - Agenda & Minutes
- Richard Norman (Unlicensed)
Date
Attendees
Richard Norman (Unlicensed) (Pistoia Alliance)
Craig Dickinson (Unlicensed) (Eli Lilly)
- Chris Lloyd (Deactivated) (MedImmune)
- Carmen Nitsche (Deactivated) (Pistoia Alliance)
- Kieran Todd (Deactivated) (GSK)
Discussion items
| Time | Item | Who | Notes |
|---|---|---|---|
| 17:00 - 18:00 (60 mins) | >Discussion on identifying ‘value’ assays >Use of sequence/structure in developability risk prediction | Richard, All | >Collated criteria for building set of ‘value’ assays >Highlight ‘gaps’ in current practices
|
| If time permits | >Wiki page and WoW >EMBL-EBI presentation outline. |
Minutes
Wiki page and Ways of Working
Link to AbVance II wiki page: https://pistoiaalliance.atlassian.net/wiki/spaces/ANT/overview
>Wiki will not replace Google Docs. Wiki is for collaborative work and can link to documents (e.g. Google Docs) if we so wish. It is up to the team to decide what works for everyone.
Discussion on number of organisations and expertise required
>Working towards getting representation from other organisations, a critical mass of 10-12 is seen as ideal.
>With regards expertise, ideally need one person who has good oversight of what that organisation does in developability space but whether they come from research or development doesn’t matter. More than one person from the same organisation can be on the team.
>Depending on how structure/modelling strategy evolves as part of the project may need to involve people with the relevant expertise.
>We need line of sight to appropriate management level within each organisation which will provide sign-off on data sharing. The proposal requires their input and approval sooner rather than later.
Identifying ‘value’ assays
>Agreement with approach for creating a database of 'value' assays and associated data (for standard mAbs):
- Agree important criteria (associated with developability risk) e.g. part of TPP
- Identify biophysical assays which provide relevant information
- Select which assays provide most 'valuable' information
- Share SOPs for these assays and guideline metrics (data ranges for assay output)
- Test, validate and show correlation with endpoint data (where this has been established)
- Generate standard assay data for selected (public or internal from 'dead' projects) molecules
- Test if correlation holds for non-standard mAbs
>Need to check what level of detail companies would be willing to share e.g. from presentations given at external meetings there is evidence that companies may not be prepared to give away composition of assay buffers. Without this there would be minimum value in database, as such we need to define what is data needed to provide minimum value.
>At this stage it is difficult to predict how many molecules and how much associated data will be available and whether this will be enough to have an impact.
Use of sequence/structure
>Agreement that don’t need to share sequence/structure as can use e.g. structural descriptors and public and/or data from ‘dead’ projects as alternatives. In addition, could use existing published algorithms which derive scores from structure data.
>Need more thinking on below approach to using sequence/structure to predict endpoint data:
- Establish current internal practices for use of sequence/structure information (what has been done internally?)
- Highlight 'gaps' which could be addressed through, more data, better data or better understanding (i.e. correlations)
- Combine sequence/structure data with standard ‘value’ assay data
- Seek correlations between sequence/structure and assay data > assay data and endpoint data
Challenges (risks)
Potential challenge: internal push back to sharing assay data correlation with endpoint data (especially if not equitable return from partners). Potential solution: pick 1-3 assays for which data is being shared and establish their correlation with PK, viscosity, solubility etc. internally, i.e. build model which can be shared.
----------
Richard Norman – 7 May 2018