tech

AI's human protein database a 'great leap' for research

8 Comments
By Patrick GALEY

The requested article has expired, and is no longer available. Any related articles, and user comments are shown below.

© 2021 AFP

©2024 GPlusMedia Inc.

8 Comments
Login to comment

I’m rarely one to wave my country’s flag - our football fans are a national disgrace - but Demis Hassabis, the brains behind Deepmind, is the opposite: a once-in-a-generation mind with the will and drive to put it to good use. What he and his team have done here is infinitely more significant than a bunch of meatheads winning a bunch of medals.

3 ( +3 / -0 )

Again... ML (machine learning), not AI (artificial intelligence). The key difference being that the former currently exists and can be used, while the latter does not and can not.

Machine learning is a subset of AI.

1 ( +1 / -0 )

If you've ever tried to understand protein folding, beta sheets, alpha helices, what all, just trying to get your mind around one protein was a brain popper. But, now, the real task begins as we look for what co-ordinates and operates and determines the 'on' or 'off' of ALL of these many proteins from conception to senescence. This may not be quite as easy (relatively) because much of what we will need to look at has previously been labeled "junk" DNA (circa 1980) by serious "meatheads" in the biology community. This control structure is as old as life itself and almost certainly highly conserved and we see in bacteria that the same stretch of DNA can produce several products by just, say, shifting the reading frame one base and a whole new functional product results. We see there a level of information integration that is stunning on its own but, when looking at the extent of so-called junk DNA in the Human genome and assuming that, perhaps, the same 'multiplexing' of information occurs there but is expressed as small RNAs, we see that we've still only just begun to unravel how we work. Maybe next week another team of 'geniuses' will announce a now unimaginable technique for mapping the control system of a eukaryotic cell but, failing such a development, it will still be a long time before we understand the cell. And for those 20K+ proteins, each may have several alleles in the collective genepool and each of those may have slightly different effects alone or when combined with other alleles. Still, "great leap' is an apt description of this advance. I wish I could be around to see how it works out.

0 ( +0 / -0 )

Again... ML (machine learning), not AI (artificial intelligence). The key difference being that the former currently exists and can be used, while the latter does not and can not.

0 ( +1 / -1 )

Huge news and a really important advancement not only because of what was done but also because it validates a method to keep doing it. A lot of these predictions may be inexact right now (we still don't know the full extent of the factors that modify the folding of the proteins) but each new discovery will help making these predictions more and more precise without having to do everything "by hand".

And even now, a lot of work right now is done by searching relationships between proteins and other molecules, so if a huge lot of structures are known this search becomes much more productive.

0 ( +0 / -0 )

RecklessToday 04:05 pm JST

Very impressive. I hope it is put to good use and not to develop biological weapons that will destroy us.

AMEN! You called that one right on Humanity's nose! THAT is a #1 concern for ANY advance we make in almost ANY area of new knowledge. What will the monsters do with this? I guess we'll find out...

0 ( +0 / -0 )

'Artificial intelligence' versus 'machine learning'? How about "artificial learning' or 'machine intelligence'? What difference does it make how someone who doesn't understand it anyway calls it? And has someone finally fully defined 'thinking' yet or, better, 'intelligence' or even 'artificial' in this context? Words, both our most vital tools and worst nightmares. What is MOST important is what comes out of it in behavior...Skinner Lives!

0 ( +0 / -0 )

Login to leave a comment

Facebook users

Use your Facebook account to login or register with JapanToday. By doing so, you will also receive an email inviting you to receive our news alerts.

Facebook Connect

Login with your JapanToday account

User registration

Articles, Offers & Useful Resources

A mix of what's trending on our other sites