We've discovered more about the world
than any other civilization before us. But we have been stuck
on this one problem. How do proteins fold up? How do proteins go from a string
of amino acids to a compact shape that acts as a machine and drives life? When you find out about
proteins it is very exciting. You can think of them as
little biological nanomachines. They are essentially the
fundamental building blocks that power everything
living on this planet. If we can reliably predict
protein structures using AI, that could change the way
we understand the natural world. Protein folding is one of these
holy grail type problems in biology. We've always hypothesized
that AI should be helpful to make these kinds of big scientific
breakthroughs more quickly. And I'll probably be looking at little
tunings that might make a difference. It should be creating a distagram one
and a background distagram one.
We've been working on our system AlphaFold
really hard now for over two years. Rather than having to do
painstaking experiments, in the future biologists might be able
to instead rely on AI methods to directly predict structures
quickly and efficiently. Generally speaking, biologists tend to be
quite skeptical of computational work. And I think that skepticism
is healthy and I respect it, but I feel very excited about
what AlphaFold can achieve. CASP (Critical Assessment of Protein Structure Prediction) is when we say,
look DeepMind is doing protein folding. This is how good we are. Maybe it's better than everyone else,
maybe it isn't. We decided to enter the CASP competition because it represented
the Olympics of protein folding. CASP, we started to try and speed up the
solution to the protein folding problem. When we started CASP in 1994, I certainly was naive about
how hard this was going to be.
It was very cumbersome to do that
because it took a long time. Let's see, what are we
doing still to improve? Typically a hundred different groups from
around the world participate in CASP. And we take a set of a hundred proteins, and we ask the groups to send us
what they think the structures look like. We can reach 57.9 GDT
on CASP12 ground truth. CASP has a metric on which you
will be scored, which is this GDT metric. On a scale of zero to a hundred,
you would expect a GDT over 90 to be a solution
to the problem. If we do achieve this,
this has incredible medical relevance.
The implications are immense. From how diseases progress. How you can discover new drugs.
It's endless. Well I wanted to make
a really simple system, and the results have
been surprisingly good. The team got some results
with a new technique. Not only is it more accurate,
but it's much faster than the old system. I think we'll substantially exceed
what we're doing right now. Well, then it's a game changer, I think. In CASP13, something very significant
had happened. For the first time we saw the effective
application of artificial intelligence. We've advanced the state of the art
in the field, so that's fantastic, but we still got a long way
to go before we've solved it. The shapes were now approximately
correct for many of the proteins but the details,
exactly where each atom sits, which is really what one would
call a solution were not yet there. It doesn't help if you have the tallest
ladder when you're going to the moon.
We hit a little bit of a brick wall,
since we won CASP, then it was back to the drawing board
and like what are our new ideas. And then it's taken them
a little while, I would say for them to get to where they were
but with the new ideas. And then now I think we're seeing
the benefits of the new ideas. They can go further, right? So that's a really important moment. I've seen that moment so many times now.
I know what that means now. And I know this is the time
now to press. We need to double down and go
as fast as possible from here. I think we've got no time to lose. So the intention is to enter CASP again. CASP is deeply stressful. There's something weird
going on with the learning, because It is learning something
that's correlated with GDT, but it's not calibrated. I feel slightly uncomfortable.
We should be learning this you know in the blink of an eye.
The technology advancing outside
DeepMind is also doing incredible work. There's always the possibility
another team has come somewhere out of left field
that we don't even know about. Someone asked me,
"Well, should we panic now?" Of course, we should
have been panicking before. It does seem to do better,
but still doesn't do quite as well as the best model. So it looks like there's room
There's always a risk
that you've missed something, and that's why blind assessments
like CASP are so important to validate whether our results are real. Obviously, I'm excited
to see how CASP14 goes. My expectation is we get
our heads down, we focus on the full goal which is to solve
the whole problem. The Prime Minister has announced
the most drastic limits to our lives that the UK has
ever seen in living memory. I must give the British people
a very simple instruction, you must stay at home. We were prepared for CASP
to start on April 15th, because that's when it was
originally scheduled to start. And it's been delayed
by a month due to coronavirus. I really miss everyone. And I struggled a little bit just
kind of getting into a routine especially with my wife,
she came down with the virus. I mean, luckily it didn't turn
out to be serious.
CASP started on Monday. Can I just check this diagram
you've got here, John. This one where we ask ground truth.
Is this one we've done badly on? We're actually quite good on this region. If you imagine that
we hadn't had said it came around this way,
but had put it in the right spot. Yeah, there instead, yeah. One of the hardest proteins
we've gotten in CASP thus far is a SARS-CoV-2 protein called ORF8. ORF8 is a coronavirus protein. We tried really hard
to improve our prediction. Like really, really hard.
Probably the most time we have ever spent on a single target. So we're about two-thirds
of the way through CASP, and we've gotten three answers back. We now have a ground truth for ORF8,
which is one of the coronavirus proteins. And it turns out we did
really well in predicting that.
It's an amazing job everyone,
the whole team. It's been an incredible effort. Here what we saw in CASP14 was a group
delivering atomic accuracy off the bat. Essentially solving what
in our world is two problems. How do you look to find
the right solution, and then how do you recognize you've got
the right solution when you're there. All right. Are we mostly here? I'm going to read an email. I got this from John Moult. And I'll just read it. It says,
"John, as I expect you know, your group has performed
amazingly well in CASP14. Both relative to other groups
and in absolute model accuracy. Congratulations on this work.
It is really outstanding." AlphaFold represents a huge leap
forward that I hope will really accelerate drug discovery
and help us to better understand disease.
It's pretty mind blowing. You know these results were for me,
having worked on this problem so long after many, many stops and starts
and will this ever get there, suddenly this is a solution. We've solved the problem. This gives you such excitement
about the way science works. About how you can never see exactly or even approximately,
what's going to happen next. There are always these surprises.
And that really as a scientist
is what keeps you going. What's going to be the next surprise?.