Sam Fazeli

to be able to put a terrible look at?

So that's, again, we've talked about cell, we've talked about networks, we've talked about protein targets.

On the proteins, in terms of developing, perhaps, either being able to use the modified proteins as therapeutics themselves, but specifically in that case, how do you avoid the problem of immunogenicity?

Because that is, the further you get away from what I would call natural biology, try and modify something,

the higher the risk that you'll have that.

Or do you not believe that?

Okay, so I'm going to make an assertion here or a statement for you to tear apart.

Application of AI is going to be more fruitful in the hands of a large pharma company than a biotech.

Why?

Large pharma companies have, at least in certain therapeutics areas where they've been gathering data and collecting information, generating data, lab notes, et cetera, for decades, assuming that data is going to...

be available to interrogation by AI, i.e.

it's been correctly annotated, it's already in the right place, they should have a better chance of using AI in terms of training AI for their models than a smaller startup or biotech company, etc.

That's the statement.

What's your reaction?

Okay, that was a perfect answer.

Something that keeps going around my head and I keep asserting that point and I'll stand corrected.

Talking about data, one of the problems for training models on public data is that there is a risk that negative data, experiments that didn't work, are not published.

Now, this is not my great idea.

I think a lot of people have raised this issue.

How do you deal with that when you're training an AI on, say, everything that's ever been published on PubMed or PubChem or whatever?