Boehringer Ingelheim swaps lab coats for AI algorithms in search for new drugs

Boehringer Ingelheim is the latest pharma company to turn to AI in the hunt for new treatments and therapies. The German company announced earlier this week that it is working with IBM to use Big Blue's foundation model tech "to discover novel candidate antibodies for the development of efficient therapeutics." The plan is to …

[Roland6]

Statistical analysis rebadged as AI ?

This just seems to be traditional statistical analysis upgraded to use a modern large model engine (ie. More parameters and thus richer associations than previous generations) and sprinkled with AI fairy dust.

Not saying they haven’t improved their chances of finding new compounds, to feed into drug research, but this hyping of “AI”…

I once did some statistical modelling with Excel :-O (and prior to that some constraints-based/rules-based systems programming); Perhaps I should add “experienced AI programmer” to my CV.

[Spoobistle]

No less lab coats

The use of AI for generation of new candidate therapeutics is unlikely of itself to be a threat - this is the start of the process before all the effectiveness and safety testing. Most of the previous approaches have the same problem that they generate many candidates from which you then have to pick the winners. It's not like Pharma hasn't done that before (and we all know what happens when it goes wrong) so I'd be more worried about AI being used to interpret effectiveness or safety testing data.

The antibodies slant does seem to be a novelty - most drugs are "small" molecules that fit like a key in a lock to disrupt some process. Antibodies are much bigger and work the other way round, by enveloping themselves round or sticking to a smaller entity (epitope). Historically, drugs were made by chemists and antibodies by biologists - it's only relatively recently that antibodies have been recognised as potential disease fighting molecules*. The classical way of making an antibody is by immunising a person, or animal. This is a very hit and miss process but its mechanisms are becoming better understood through systems like Alpha Fold, so I can see why AI would be a good bet to move the field on in a rational direction.

It's not going to reduce lab coats though - still got to make and test the things!

* Though anti-toxins and anti-venoms have been used for specific purposes for about a century.