...
Discussion items
| Time | Item | Who | Output |
|---|---|---|---|
| 17:00 - 18:00 (60 mins) | >Database questions >Discussion on what internal or collaborative efforts are ongoing which could 'compete', 'overlap' or in some ways 'dilute' the value and interest for what we are proposing and working towards? >Next steps discussion: focus on our Approach | >Laura to introduce >All >All | >Agree approach to database setup discussion. >Capture efforts and agree approach to differentiate or combine if feasible. >Team input on proposed Approach. |
| If time permits | >EMBL-EBI presentation | >Team input on proposed presentation. |
Minutes
Internal collaborative efforts
>Comments from meeting or sent through by email:
>Francisco (Abvasnce Biotech): Even though it is a truism, pooling intellectual, material and procedural resources from across industry can enable us to accomplish goals that would stay out of reach of individual organisations. This is especially true of pseudo-exploratory campaigns where (ideally) lots of (heterogeneous) data must be examined, validated, compared and used to support or dispel hypotheses. Beyond that, a correct assessment of the intrinsic sources of errors and variability across organisations cannot be adequately inferred from publications or meetings alone, requiring instead a focused and coordinated effort to compare assays, conditions and selected outcomes.
Finally, there is the aspect of identifying the most suitable minimal set of assays that can "do the job" of weeding out bad from good antibodies while still conserving full predictive power. This type of task is best accomplished within a diverse group willing to share different experiences (or, at least, this is my current understanding).
>Chris (MedI): Internally, we periodically review (and refine if necessary) our early stage screening assays based on experience of how the molecules ultimately behave in Development i.e. whether problems were predicted or not. So in some way, I get the point of the guys from Pfizer, in that individually we’re likely to get close to the minimal set of most predictive assays at some point ourselves. However, Francisco also makes a good point, that working together would make this more efficient and may uncover something we wouldn’t as individuals.
So I do see value in doing this across different organisations, but potentially what would be greater will be combining data / experiences so we can learn something about what makes “good” versus “bad” antibodies based on their physicochemical and structural properties.
>Bryan (Lilly): In principle (not resourced yet) Lilly are developing an internal strategy to link early stage assays to endpoint assays. Using combination of experimental and computational tools and holistically assessing all. Pistoia Alliance AbVance II efforts and internal Lilly efforts are seen as complementary to each other. The specific value offered by AbVance II is specifically around (1) getting an external perspective, (2) access to the number and diversity of molecules on offer, and (3) shared workload vs shared benefit.
>Laura (GSK): Have conducted an internal exercise looking at the predictability of late phase assays based on data from early phase assays - focus was on predicting aggregation from high throughput assays with low concentration of material to predict outcome of later stage assays with higher concentrations but smaller number of samples - work has been completed and currently working on collecting more data based on output. The specific value offered by AbVance II is in having access to more datapoints on which to build models. Experience has shown that more data is needed than what is currently available internally.
Approach workflow
>Agreement that team will capture comments on our 'Approach' wiki page offline. Once this has been done we can tackle this item at the next PTM.
Database questions discussion
>Link to questions posed by Laura and Kieran: Outstanding Questions
>These originated from trying to predict which questions internal stakeholders are likely to ask and were aimed at helping all to obtain more clarity on what data can be shared.
>The team agreed that we need to have a clear view of what we want when engaging internal stakeholders and push for this rather than find out what we can get. Our 'want' is in the form of a BASECASE, which is the minimum information we feel needs to be shared to meet some of our objectives and which in turn will deliver the value proposition. In parallel to developing our BASECASE we should implement a TIERED approach to understanding what we are likely to be able to share openly, what we are not going to be able to share at all and what lies in between. This can be done using the following categories as an example:
- what could we do if it was all open?
- what could we do if it is only shared with members?
- Would it make a difference if the data was anomymized?
- what could we do if we don't share the data but just the models built on the data?
>Marrying the BASECASE with the TIERED approach will form the basis for the proposal for stakeholders.
>The engagement with stakeholders will provide feedback on the proposal and a sense of the requirements to be worked into the proposal going forward.
>Pistoia Alliance has the legal framework in place to cover the eventualities of data sharing.
>The team agreed to populate assay names to the best of our ability in the BASECASE table over the coming week.