In the Heat of Sevilla

With temperatures close to 40°C during the day, Sevilla certainly provided an appropriate setting for this year’s BEBPA European Bioassay Conference.
After attending the BEBPA US conference in March, I had probably been a little spoiled from a statistician’s point of view. There, the program offered quite a few talks with a strong statistical focus. In Sevilla, the statistical yield was noticeably smaller.
30 September, 2026 / Dr. Andreas Fromkorth
Artificial Intelligence

AI takes center stage

Instead, another topic is becoming increasingly prominent — perhaps unsurprisingly: artificial intelligence.
So far, I am not entirely convinced by everything I have seen in this area. One presentation, for example, looked at using different large language models to directly analyze plate effects. It was certainly interesting to see what the models came up with, but my immediate reaction was a rather practical one: how would you ever validate such an approach?
For exploratory work, LLMs can obviously be useful. But once they become part of a regulated analytical process, questions around reproducibility, transparency and validation become much harder to answer. At least at the moment, I find it difficult to see how a method like this could be brought into a properly validated workflow.
Workshop insights

Back to bioassay basics

My personal highlight of the conference was much more traditional: the Basic Bioassay Workshop on the second day.
Mike Merges and Anton Stetsenko — both long-standing members of the BEBPA community — once again gave a very confident and entertaining introduction to the world of biological assays. The workshop is offered at many BEBPA conferences, and this was not my first time attending it. Still, I learned something new.
The Freezer Door Effect

Variability in unexpected places

One of the examples concerned a source of variability that had never really been on my radar before: repeatedly opening and closing the freezers in which bioassay materials are stored. Maintaining constant storage conditions can apparently be much more difficult than it sounds.
I found this particularly interesting because, as statisticians, we often talk about sources of variability in fairly abstract terms: analyst, day, plate, instrument, reagent lot and so on. It is easy to forget that behind these factors there are very practical laboratory processes. Sometimes a source of variability really can be as simple as someone opening a freezer door too often.
Getting in contact

Insights from the booth

Our booth was also well attended throughout the conference. As is often the case at BEBPA events, many of the people stopping by were already PLA 3.0 users. That does not make the conversations any less valuable. Quite the opposite: it is always useful to hear how PLA 3.0 is being used in practice, which questions come up in day-to-day work and where users see room for improvement.
Last-Minute Reading Material

An update from USP

The biggest piece of news, however, reached us when the conference was almost over.
We were already dismantling the booth when Sonja Klingelhöfer from Richter BioLogics pointed out that USP had just published the final versions of the revised General Chapters <1032> and <1034>.
That immediately gave us some new reading material for the journey home.
For us, <1034> is of course particularly interesting, as it deals directly with the analysis of biological assays. Together with the revised <1032>, it will certainly keep the bioassay community busy for a while.
So while Sevilla may not have delivered quite as much statistics as I had hoped for, it still provided plenty to think about — from AI and freezer doors to new USP chapters.
Bioassays remain an interesting field. And apparently there is always another source of variability waiting somewhere.