I. Welcome to ID the Future. I'm Eric Anderson and today we'll listen to the second half of my

conversation with Dr. Paul Nelson about the widely published announcement that Google's

deep-mind Alpha Fold had solved one of the longest standing problems in biology, the protein folding problem.

Last time we discussed what Alpha Fold is actually doing, operating is a sophisticated comparator rather than a predictor from first principles,

and why this limits Alpha Folds ability to handle unfamiliar sequences.

Now we'll pick up the conversation as I ask Dr. Nelson about these isolated regions.

So you had mentioned something to me earlier about what you called the isolated region of sequence space.

Is that related to this?

Totally.

So, sequence space is an abstract notion that is fairly easy to visualize.

Again, you can do it with a natural language parallel but let me just do it this way

we've got an alphabet of 22 depending on how you count let's say 20

20 amino acids and let's say you have a protein that's actually a rather

small protein of 100 amino acids in length. So again, beads on a wire, 100 beads on a wire. Now you say, well, how many different

hundred amino acid proteins might there be. Well that is 20 raised to the power of a hundred.

All right that is a huge space yeah And a hundred is a short protein. There are proteins that are very much longer than that. I think

Yeah. We're picking a kind of a dinky protein when you say 100 amino acids in length.

So sequence space is an abstract notion representing all the possible variants that could

exist using a 20 character alphabet where you've got, again,

I'll go back to natural language, let's say you've got a sentence of a hundred characters in length and you know including punctuation

and so forth and you've got a 26 character alphabet in English.

Well, you can imagine that could be,

...