“There is no one overall strategy for developing or implementing AI in or across the the UK.”
New advancements in artificial intelligence have the capacity to deliver major breakthroughs in cancer therapy. AI-powered protein design company Profluent used AI to make an open-source gene editor called OpenCRISPR-1. OpenCRISPR-1 can create molecules with the power to edit the human genome, according to Profluent. Profluent’s aim is to develop gene editors that are more operative than existing biological mechanisms, especially mechanisms that prevent illnesses like cancer and fight viruses. AI presenter Priya Lakhani joins The Context to explain this development in health and technology.
you’re watching the context it’s time for our weekly segment AI [Music] decoded yes it’s our regular weekly appointment with all things artificial intelligence if you’ve watched this series before you will know by now that in each of our programs we like to focus on a particular theme last time out we looked at the different approaches to regulation in the United States and here in Europe this week we’re going to focus on health you know doubt have read that AI is already delivering huge breakthroughs in medical science well tonight we will show you how the technology is being used to deliver enormous advances in cancer therapy it’s about improving automation improving efficiency improving the confidence to deliver higher doses in fewer fractions so we can reduce waiting list which is particularly important in the UK right now but how much investment is there in this new technology and are we taking full advantage a report this week from digital Health that the government has slashed in half the investment it is putting into the NHS AI lab we will speak tonight to the former NHS director of AI Innovation who these days is using the AI to streamline regulations so our Public Services can more quickly make use of what is’s on offer so let me set the context for you currently here in the UK there are are 7.5 million people waiting for treatment in England and in spite of all the best efforts it has come down from a peak of nearly 7.8 Million last September but clearly not far enough now we spend in 2022 23 the last year of confirmed NHS spending in England over 180 billion spending has increased by an average of 5.5% a year here in real terms since 201920 so it stands to reason that we need a better solution more spending does not necessarily deliver better outcomes and that is where AI promises so much our resident expert priia Lani AI educator CEO of century Tech is here with us so where have you been so I’ve only been to gwick is that it I gave you a budget you went to gwick yeah it wasn’t much of a budget but uh so I went to somewhere in England and I went to a place called Electa a company and they focus on radiotherapy and it was an absolutely extraordinary demonstration of the technology that they’ve built because we’re AI decoded I want to focus a little bit on the AI and what you’re going to see is artificial intelligence being used to be able to track a tumor in the body if you think about prostate cancer Christian uh when a patient has prostate the tumor moves right and so you need to be able to track the tumor while while you’re treating it because otherwise if you think about it you’d have to give quite a margin of error and you’d potentially be treating healthy tissue so imagine technology that can use live MRI scanning that you’re about to see track the tumor movement and then as you saw in that headline there potentially take the time to treat a patent patient down and therefore have a huge impact on waiting lists okay super exciting let’s have a look one in two people get cancer in their lifetime and around half of those will be treated by radiotherapy as part of their treatment I’m here at Electa to understand how radiation therapy works and look at their gamechanging AI technology so D you’re the managing director of Electa can you show me how conventional radio therapy works and what you’ve delivered right I’m going to jump on this machine just mind your head as you go back down so imagine if you will you’re the patient the first thing we do is make sure that you’re lying in the same position every day and you can stay still pretty well what we’ve got here is a linear accelerator and we electrically generate a therapeutic beam of radiation that comes out here now what’s important is that when the radiation beam comes out it needs to be shaped so in the radiation head here we have something called a multi-leaf cator which comprises 160 little metal leaves that move in real time to shape to The Irregular tumor volume so you’d be set up under the machine we’ve got traditional KV Imaging on this system and as the machine is treating you can image and the Machine moves around the patient to deliver the therapeutic dose effectively painting it to the tumor volume so D we know that patients really try and stay still on these machines but the tumor moves inside how have you developed Cutting Edge AI technology to be able to solve that problem and have this Precision been not affect healthy tissue this is soft tissue Imaging so what we have here is an MRI system which we’ve developed in collaboration with our clinical Partners to combine with state-of-the-art delivery system as I explained in order to get soft tissue Imaging at time of treatment combining these two technologies has enabled us to precisely deliver shaping the beam very precisely at the same time as using MRI imaging to get soft tissue Crystal Clear Vision of the anatomy in real time what I’d like to show you is what we’re doing with that information so here you can see a long volume so the dark area here is the long what the MRI imaging does is real time in real time is take soft tissue images and reconstruct them in three dimensions so you’re looking here at movement in all Dimensions In and Out side to side backwards and forwards the machine learning algorithm and where this really comes into play is as a training phase whereby the patient is set up on the system and the the algorithm is watching the delineated Target volume that’s the red bit here that the doctors created in the treatment plan yeah that’s where you want your radiation to deposit all of its dose you want to avoid the healthy tissue so the template that’s been generated by the algorithm using machine learning is delineated in blue as I’ve said and what you’re seeing along here at the bottom is a little trace of the movement in three dimensions of that Target volume as soon as the target volume volume that you want to radiate goes outside of that preset window you can see that radiation beam is automatically turned off are we going to be solely relying 100% on the AI when it comes to the treatment or is there some sort of safety mechanism that we have if something goes wrong that’s a great question there are always trained operators sitting at this control desk um they’re the radiographers they can be the physicist depending on the complexity of the treatment so you don’t 100 % rely on it but what’s so cool is that from an efficiency point of view I showed you the setup in the other room where we set the patient up you can automate really large parts of the workflow and that’s our intention in the future what is the actual impact in terms of treatment times with this technology well one of the most exciting um impacts is the confidence that it gives to the clinician and the people delivering the treatment because if you can know absolutely where your radiation beam is painting the dose in real time you can have the confidence to shrink the margin around the target traditionally radiation therapy has margins around the target margins of safety margins of error if you can shrink that you can reduce side effects interestingly as well if you have confidence about where the radiation beam is going you can escalate the dose so imagine for example prostate treatments that previously were treated in upwards of 30 treatments every day imagine with that level of Technology available to you and that confidence you could actually reduce the number of treatments down to five or even two what is the future of using AI technology in radiation therapy I think it’s about improving automation improving efficiency improving the confidence to deliver higher doses in fewer fractions so we can reduce waiting list which is particularly important in the UK right now but also across the world because we want to give hope to everyone with cancer and the only way you can do that is to provide access to care mind blown mind blown I mean that’s it that is tracking in real time increasing dosage in real time cutting the number of appointments yeah you can change the radiotherapy plan daily right because you have a visual on how the tumor is also reacting to that Precision beam so so if you’re in the UK and we had that Really Brave lady just a couple of days ago say to the Prime Minister say to cus dama is the NHS broken and she talked about her treatment and how a friend of hers lost their life during being on these waiting lists so the impact is absolutely huge chistian in terms of you go through so many more patients in such quick a time yeah cutting the weight in this the problem is as you said regulation what a lot of people out there won’t know is that you helped co-write the regulation the review for AI in healthcare with Chief the chief scientific the former Chief Scientific Advisor that’s right so last year the chancellor asked us to review regulation of AI across all sectors one of the most interesting points and I’m looking forward to having Joseph come on talking specifically about the NHS but one of the specific points that we actually learned was that because we don’t actually have overall regulation of AI and they’re going to do this on a case-by casee basis in the UK one of the issues we have to solve is that Regulators themselves aren’t actually upskilled in this area to be able to think is this but also the incentive isn’t there so if you’re a regular you’re there to keep people safe aren’t you so you’re trying to drisk so now we’re asking them to innovate and can you see how these two things conflict we’ve got a lot of work to do in the UK to be able to take advantage of this and I know that Joseph is going to shed some light on why the NHS specifically might be struggling to do this at the moment well five years ago the the then Health secretary Matt hanock launched something called the NHS AI lab there was an initial 250 million pound investment to work on challenges in healthcare including early cancer detection new denture treatments more personalized care but reports this week are that just as the technology leaps forward the budget is being slashed so let’s bring in then Joseph Connor who was previously the NHS director of AI Innovation he’s now the CEO of careful AI limited we’ll talk Joseph about what you’re doing to streamline the process very shortly but maybe you could pick up what Pria was just saying about the challenges you face when you were in the NHS when it came to adopting new Ai and and the technology okay so if we go back to when that was which is back 2027 2018 the challenges at that time weren’t just the issue of Regulation it was it was a relatively new area of awareness for the NHS so there’s quite a lot of difficulty for people to understand what it is and there is to a large degree a lot of confusion as to their value but when you raise um awareness of case studies like you’ve just done with the LCT it becomes obvious that there’s a role for things that can provide uh Predictive Analytics about what technologies can be used and applied to healthcare I think one of the challenges that I faced at that time which is we were trying to get more people in the NHS to actually be The Originators of of the AI uh it came down to the issue of resources um the air lab was just about starting there wasn’t exact any anybody to that a a clinician could turn to to say I want to address this issue can you find me somebody either internally to the NHS or externally to the NHS to help me so the NHS AI lab was formed and its job was primarily to try and bring AI into the NHS and if you think about the degree of investment that was provided it was primarily to help people get a foothold into the NHS and create the evidence to show that things were clinically safe it wasn’t to actually develop the product um however where we are at the moment is there has been a cut um but there’s still a focus I believe within the NHS and its relationship with the MH to keep regulation to a minimum but whether they’ve got enough resource to be able to do that is questionable one of the challenges that exists is it takes maybe 6 to 12 months just to get what’s called a class one device a device that just gathers information together and it can take between two and five years to get a class two device something that gives clinical advice like the one you just shown now that’s two to 5 million pounds just to get through the door so Jose and to deal with all the regulation Etc and Joseph do we have the right product managers in the AHS and if you for example can sample one of your AI pieces of software in one trust how transferable then is that to another NHS trust because I know from some of the meetings that I’ve been in the past actually it’s pretty fragmented I’ve heard in terms of how the NHS can work okay I think one of the challenges is believing that the NHS is one single body it’s the largest or sixth largest organization in the world with millions of people in it but in essence it operates in uh regionals uh where you’ve got people who are individually responsible within their location for commissioning and by commissioning I’m talking about buying technology and services so there is no one overall strategy for developing or implementing AI in or across the the UK so so what are you do what are you doing Joseph in your in your new role and with the experience you have of that what are you able to do to change the system okay well one of the challenges and arguably the biggest challenge is uh Gathering the evidence to show that your system is safe and it’s fit for purpose and to do that you need typically need to gather evidence using people and research and that’s a long process so what we’ve developed is a series of AI agents that can gather information to enable you to present that to the relevant authorities we’ you’re using AI to deploy the AI yes right I mean it sounds strange but you can imagine a scenario I won’t give you exact scenario so if you’re de developing Ai and you want it to be fit for purpose but you need to prove that you’ve engaged the audience and that the the AI is appropriate for the audience that you want to apply it to to do that you need to understand what the needs and aspirations and the health outcomes are of people in essence you need to interview them so what we do is use a chatboard environment to Mo use motivational intervie to ask people if they were to use this technology what do they think are they going to be the outcomes are they happy to use it what are there going to be the implications for them what biases that they have so it’s Gathering specific information about where they are and what they want that’s important it’s a requirement for that in for in regulation Joseph very quickly you sit on the standards committee when it comes to Ai and Healthcare so do Britain need to develop its own standards or do we need to have aligned standards with the EU in the US so that companies that are developing all of this technology have a huge Market that they can access um you how do we how do we go about this so that we’re not stifling Innovation and actually trying to leverage AI in Britain and take advantage of the opportunities of not just technology built here but also elsewhere the important thing as far as the supply is concerned is that they’re able to comply with legislation in the countries that they want to operate in if we have different legislation in the UK to America or to Europe it’s going to be very difficult for people to come to the UK and set up and say I want to operate here because and I want to develop the marketplace here because it offers me market potential there is some variance in the standards that we use in the UK to what happens in Europe and what happens in America and it’s very very important to make sure that those are aligned if we don’t do that why should people come to the UK to develop products why should they Implement products in the UK and there are people in government and some of the work that you’ve done has highlighted the need to make sure that that is happening but I can’t emphasize it enough it just needs to happen yeah need to be able to be bu that Joseph it’s a really good introduction to uh something we’re going to talk about after the break but thank you very much indeed for that Joseph Connor there so that is an insight into how we tackle disease using Ai and what we need to get past the regulation but what if we could prevent the disease from ever appearing in the first place coming up a company in San Francisco who are developing an AI abled protein to redesign our problematic DNA I wish you could hear what priia talks about in the break she blows my mind with some of these things it’s not burnly football it’s not Burnley football it’s it’s the moonshots like we’re going to talk about now because 10 years ago some of the leading geneticists in the world developed a technology to modify DNA sequences within living organisms it’s called crisper you’ll have seen reports before about the potential application of that Tech technology in targeting genetic disorders but that technology was using natural proteins we’re going to introduce you to a company in San Francisco that’s just developed synthetic molecules designed and controlled by AI they were developed and styled on natural proteins there are something like 5.1 million of them that the AI model learned to recognize this new synthetic molecule is able to Target specific DNA sequences where they can edit out out potential diseases and conditions here uh with us uh from uh the company Ali madani he is the CEO of profluent um did I did I explain that correctly Ally was that was that the right sort of explanation of what you’re doing Christian that was that was absolutely fantastic um this is really a scientific moonshot that there’s a lot to unpack here specifically um we really sought out to learn the language of biology the underlying blueprint that nature has provided us so that we can design novel medicines breakthrough medicines from scratch using AI as our interpreter and our writer and designer that’s correct and so I I don’t really know how Christian feels about using AI to modify or redesign the genetic code of humans um I did just want to zoom in on his face when I said that just to see what he was thinking but Ali you have demonstrated that you can edit or rewrite the human Human Genome with I can you please unpack that for us and our wonderful viewers absolutely um it really starts with DNA in particular uh where DNA is our fundamental blueprint of life and human health as well um and there’s a wide range of diseases that have determinant that would be found in DNA so one example of which would be the CLE cell disease that’s prevalent in a serious condition uh worldwide in particular um and we’ve seen really exciting approvals both the UK and the US for a crisper medicine that can take a patient um that is otherwise unhealthy which has sickled red blood cells specifically edit its uh the patient’s uh genes in particular to find a oneandone lasting cure to make the patient now healthy and not symptomatic at all um so I think that’s really the promise of Gene editing that has been enabled by crisper uh and we’re really excited to build upon that and go beyond that alog together I mean is a that is a complete moonshot we as I described it at the beginning I don’t know how much or how far you can go with this but I can well guess what the implications are I mean what other potential cures are we talking about absolutely at least in the realm in genetic disease you are now able to alter the human genome in effect and then be able to find precise cures to disease that um would otherwise we didn’t have the tools to develop altogether so AI inevitably is an interpreter um it’s learning the underlying rules of biology being able to design precise and safe medicines and also but just just to take a step back there is it is it looking at these 5.1 million molecules and saying okay that’s the tool I need I’m taking that molecule and I’m going to send it here to that pink thing we’ve just seen revolving in our screen the pink thing correct yes yeah so one way of thinking about this is um we are providing attributes and properties that come to mind we can feed into this Tool uh and specifically the AI model broadly speaking okay we want to go after this target there is this structural constraint that comes to mind there’s this functional task that we have in mind and the AI model can generate sequences and libraries of sequences that we can then pass on to biologists who will synthesize this and test this in human cells and specifically for open crisper or open crisper initiative in particular the rotating structure that was shown there that is one of those molecules that was the first world’s first demonstration of a molecule that can precisely go into human cells edit DNA and this molecule was generated from scratch using Ai and not found in nature uh in particular Ally what I really want to make here so you have this and please correct me if I’m wrong um so you’ve got this guide RNA that can go and Target the DNA and then it decid when you when it wants to cut you’ve got these molecular scissors if you like which is called the cast 9 which I think was the image that we just showed um with the green and the blue I love how we’re using um Christian and I using colors to explain this and it goes and cuts the DNA if you like um but what I really want to try and explain here is why profluent has been so Cutting Edge here because what you’ve done is you’ve op sourced the AI code um and this is really really because in the past we’ve relied on nature and Discovery right to be able to uh cure disease and what you’re trying to do is something very very different which is saying rather than happen upon potential Discovery to help us cure disease you’re saying let us design it can you explain that quickly We R out so if you if you remember from grade school specifically the example of penicillin particular where Alexander Fleming for example by happen stance had left out a petri dish and mold specifically to develop antibiotics um over a century to ago uh the process of drug development has maintained that Pro that uh underlying technique of by happen stamps coming across something in nature and repurposing it for human Therapeutics this is a paradigm shift where instead of uh relying on nature we can shortcut evolution alog together and for pressing needs diseases patients with real diseases we can feed this into the model and design from scratch a custom solution a bespoke solution to address that in a safe manner short cutting shortcutting Evolution Christi shortcutting Evolution Ali Pria thank you very much we are out of time just a reminder if you missed some of the program it’s there on BBC YouTube we’ll do it same time next week thanks for watching