Last week, DeepMind, an offshoot of Google, announced they had solved one of the great problems in medical science that has the potential to change the face of medicine. It’s all to do with what artificial intelligence, as opposed to the human sort, can solve. In this post I will attempt to explain in lay terms what has been achieved and why it should speed up a cure for diseases such as Covid-19 in the future. And it all started with the grand strategy game GO!
The spike protein on the Covid-19 virus
- We know about the central role of the spikey bits on the surface of the corona virus in causing all the problems
- It is a protein and it does its damage by locking on to one of our human proteins called the ACE receptor (in this case perhaps ‘Ace’ is not the best adjective!)
- Readers of this blog will recall that most of the vaccines in development (including the ones by Pfizer, Moderna and Astra Zeneca) are thus directed towards producing antibodies to this spike protein
- As well as vaccines, new drugs have been developed that block this protein including those, such as Regeneron, tried on some rather famous guinea pigs!
- Thus it makes sense that the more we know about how proteins work, the better the chances of finding successful cures for diseases like Covid-19
What do different proteins do?
This is about the proteins we make in our bodies and not the ones we eat! They are pretty amazing chemicals:
- There are around 400,000 different proteins in our bodies
- They are fundamental for virtually everything that makes our bodies strong and keeps us healthy
- Some common proteins you will have heard of are listed below:
How do different proteins do completely different things?
Let’s look at their length
- Readers of this blog may have read the post* where I described how proteins are made up of a string of amino acids like a string of beads
- There are 20 different amino acids: often referred to by the first 3 letters of their name eg ‘Ala’ for Alanine (I like that name!)
- These strings of amino acids can be very long
- The spike protein is now thought to contain around 1400 amino acid “beads”
- One human protein called Titin has 24000 amino acids
Is length everything?
- The short answer is not!
- What matters is the shape of the protein and this is where the wonders of nature come in
- Having a very long string of amino acids dangling around cells is pretty useless!
- Chunks of the strings often curl themselves up into a spiral (called a helix – ‘a’ in picture below)
- A protein can then be a sequence of spirals with straight bits in between (‘b’ in picture below)
- And it doesn’t end there!
- Further folding continues and very complicated shapes can emerge
- This is a picture of the protein haemoglobin
Why is the shape so important?
- This is the most fundamental thing as it is the shape of the protein, not its length nor which amino acids it contains, that makes it what it is
- It is the shape of the protein that is recognised by cells in our body and this recognition means that the particular protein locks on to a cell and allows it to perform its functions
- That is how our immune system works. Our immune cells recognise the shape of foreign proteins and this stirs them into action
- This is also how the Covid-19 virus uses the shape of the spike, by allowing it to be recognised and lock on to our lung cells which then causes all the damage
- If we could identify the shape of proteins as opposed to just the sequence of the amino acids, we could design drugs which in the case of Covid-19 spike would
- stop it sticking to lung cells
- increase the response of immune cells to destroy it
What determines the shape of a protein?
- Individual amino acids may have a positive or negative electric charge (like one terminal or the other on a battery). This can make them attract or repel other amino acids.
- The pattern of ‘+’ and ’-‘ amino acids in different strings will determine how they curl up into spirals etc
- Exactly what shape emerges we couldn’t predict until now
- For decades scientists have been able using complex microscopes to take pictures of different proteins and see their different shapes
- They also knew the amino acid sequence
- Scientists, though, couldn’t estimate what the shape of the protein would be even if they knew the exact sequence of amino acids. It was just too complicated to work out all the possible spirals and folding to give the final shape
- This is where AI (Artificial Intelligence) comes in!
- DeepMind, based in London, had used AI to solve how to win at the two person strategy game GO and beat any human opponent
- My guess is it could even beat Beth Harmon at chess (Queens’ Gambit TV series)!
- They have adapted that computer learning approach to look at structures of proteins that were known and their sequence of amino acids
- They then fed their AI system with the known shapes and the strings of amino acids and asked it to come up with a failsafe algorithm to work out what shape would result from a given set of strings
- They identified the correct structure in 95% of the sequences – a result previously considered unattainable by scientists around the world
- Given that sequencing a protein, such as the Covid-19 spike protein, is relatively easy, the technology should be able to be developed to work out its exact shape and then go on the attack with the bespoke design of drugs
- An amazing breakthrough or what!
PS I know this is a bit different to my normal post and perhaps of less immediate relevance to you the readers. My aim is to try and get over in an easy to understandable format the significance of this advance. If I have succeeded and you would like more stuff similar to this please click on like!!
If you would like to receive email notifications of new posts on this blog just click this link and enter your email at the bottom of the home page: https://makingsenseofcovid19withs.com