Some by some, all by all, HOMER and discovery/mining of new findings (given agentic AI)

I am exploring the concept of reverse translation (bedside to bench as reverse of bench to bedside).

How diligent analysis of available data (RWD or CT) can contribute to improved drug development.

Back in 2010s, there were posters at OHDSI on the topic of exploring (in a scientifically sound way) all possible pairs of drug and outcomes (all by all) or just limited subset of pairs (some by some).

I found this reference to HOMER (with links now pointing to omop.org page that no longer work) ( https://www.ohdsi.org/analytic-tools/homer-for-population-level-estimation/ )

(exaple of old link would be http://omop.org/sites/default/files/OMOP%20Symposium%20Day%202%20Symposium%206Nov2013%20Ryan%20vision%20full.pdf

I know there was a poster by JnJ group in 2018 (plus or minus few years) but for some reason I can’t find it. (even though I instructed agent to put together a giant CSV file with all abstracts going back of all time for OHDSI US conference). If you know the year and title of the posters (there was one and another one year later) - please post them here.

In fact in 2026 the latest glimpse I could find is https://www.ohdsi.org/2026showcase-11/

Is your team engaged in systematic exploration (think “all by all”) of RWD (or clinical trial data) and is there perhaps a public abstract about that (or you can share some description of it), I would be very curious (and OHDSI community perhaps too) in learning more.

Examples are: (systematic search for:)

  • positive side effect (drug X accidentally helps with outcome Y)
  • suspect adverse drug event (drug X has some ties to outcome Y that is novel (not published about))
  • disease X has feature Y that makes the disease less severe/more severe

Given agentic AI and things like co-scientist, that may be topics people are working on. If yes, please chime in with your response/example/publication/description.