[Music]
hello and welcome to technically speaking where scientists and Engineers come together to chat about common interests share knowledge and satisfy some curiosity I'm Antonia and I'm joined by Emma and Laura to talk about biophysics and how computer simulations and artificial intelligence can help develop drugs so to start off with Emma what's your interest in this as a physicist I just finished my masters in physics and my Master's project in itself was testing
alternatives to antibiotics in the lab and then using molecular Dynamic simulations to complement that and kind of further the research and the general idea of that was to test both in the lab and in the simulations two or three alternative sounds biotics and then use that data to kind of compare and see what worked and then to just all just test a completely new alternative in the simulations um having like backed it up with the experimental data and so I looked into because it's the Masters
in physics kind of like um although it was very biology based Masters um send other simulations and how um the different like distances between the you know Alternatives antibiotic and the membrane made it be more effective or less effective and how that was also shown in the lab work um because building a representation of this molecule can be done within a few days on the simulation but to actually build it in the lab can take um months and months and then you also have like uncontrollable
um defects that arise from lab work like purities within the samples it's always a mess so it's interesting for me to see how far can we actually push this idea of using simulations to complement the lab work will it save more time and money or how useful is it going to be and so that's kind of my background on using simulations to develop drugs that can help in like a biophysics um setting I guess now that's really interesting um you know the Practical aspects and almost like checking your
your simulation against real life but then also having to compensate for all the messiness that can come with real life experiments and and the lab and all those uncontrollable factors um so no that's really interesting and that was actually kind of why we wanted to talk about biophysics in this episode wasn't it so Laura what about yourself how's your jack of all trades working yeah my PhD was computational chemistry it was not involving drug development it
was for these um fairly complex polymers for carbon capture but I learned quite a lot about how you do the simulations and how you could apply that to other things or someone in my lab computational Lab that was working on getting proteins through membranes and doing all sorts of things with those uh and picking up on what Emma was saying about you get messy things going on in the experiments I agree completely and I was always working with like um really idealized versions of the
polymers and I did think well if that Bond hasn't formed in the same way in the lab how representative is my experiment on the computer but the experimentalist would always say to me but you're working with all these approximations of how atoms interact so how can you say that your simulations are accurate so which is best who can say depends on what you enjoy I guess or maybe how or the best of best of a bad lot maybe not a bad one maybe yeah I always do like the fact that I could see exactly
what my atoms were doing I didn't have to sort of try and figure out from what was going on with an experiment so that was always my real interest I can see precisely what's happening would you say you also prefer the simulation Emma living in the in the computer in the lab yeah I mean if the stars the simulations I can do in my bed which is always very nice um but I know I did have the problem where I had to redo a lot of lab work or like it was very like the lab work I was
doing is very dependent on external vibrations and involved like hand injections so if you're a bit shaky you had to completely start over again and so it was a lot of like repetitives driving for Perfection you know if you were more tired one day or had maybe another coffee then you shouldn't have then you're like well the lab work's gonna be a bit ruined today um and also like you said with seeing the exact picture of what's happening when something that shouldn't happen
happens in the lab it's always that's why and that is why you do repeats as a scientist but you're always like is that because of this new experiment I did or is that because I just made a mistake and I didn't clean something properly um and I feel like it's harder to like actually understand what's going on just by looking at the bulk behavior of the experiment that I did anyways so I think I did enjoy the simulations um but also it is nice to be able to sit down and do them and not
to wear gloves and goggles for in the labs all day um so maybe that's got a part to play I'm not sure yeah and the idea that a lot of the people I knew that were doing experimental work in the lab had to go in it like weird times are going at the weekend because the only time equipment was free and my simulation's like I set them up in the evening and come back in the morning and they're complete I can analyze them during the day and then set some more off overnight repeat
it gave me a lot better sort of work-life balance yeah once I had to work around like the other people using the lab equipment um where it's like you send a simulation you just get added to the queue and then automatically just runs they run the simulation for me so it was um I think a bit more enjoyable in that sense and it felt a bit more physicsy because when I was doing lab work I felt like do I have like enough lab experience to be good at this um so I feel like it was a lot of the
first few months were very trial and error which was a bit of a waste of time in the end but that happens with lab work you always waste time it's a learning experience it's awesome yeah the closest thing I can relate to in that way was when I was working on a particular project in the lab for three months and we didn't know how the interaction would happen and we just kind of hoped in theory it should work and when I was trying to read textbooks to figure out what would be uh what would happen
um the the end goal was just we just need to see if it works it doesn't matter if it theoretically works just if it works or not so I was just in the lab three months just running the same thing over and over um and yeah it does get physically tiring with that in your undergrad in chemical engineering it was a summer internship that I did in a nanotechnology company which sounds really cool but um but for me it was it was a shock to the system suddenly working nine to find
them in the lab you realize how much uh how how tiring it is to be on your feet all day and how difficult it is when you have to like think am I clean when you leave the lab and you know go go to like have a drink or whatever it was quite dehydrated oh yeah I sort of thing you get used to though I've done a lab work in like 34 degree heat because I had to because I had a deadline I mean I actually physically in chemistry love wearing these like really thick rubbery gloves
because I was dealing with a chemical that would absorb through normal latex gloves it's very uncomfortable it's another Pro of doing computational work yeah uh anyway I feel like we're getting a little bit distracted yeah so on the basics of biophysics I I kind of understand the general idea that physics is atomic level chemistry is sort of you know the ions and electrons interacting so a bit bigger looking at more like the molecule and then biology is sort of the result of molecular
interactions but then how does biophysics work with skipping the whole chemistry step how do you um how do you understand biophysics to to fit into the whole science picture I think this is a really good um question because I'm about to do a PhD in biophysics and I technically just doing my Master's in biophysics so I really should know what biophysics actually is but I think um and I looked at the definition and by definition it's the application of physics laws to understanding biological
phenomena sounds reasonable and I think that makes sense especially like maybe the start of biophysics but I think now as like physics has also like begun to advance different fields you know like in computation and Ai and like optimization that's become like quite a large area of physics um I think that has also been kind of incorporated into biology now and that's Now biophysics by definition but I think it's just using physics to model and interpret to try and understand in
advance what we know about biology and I don't think it does skip out chemistry I think it's just a same thing because I feel like I've done a lot of chemistry under the name of physics or under the name of biophysics because I don't know what actually chemistry is anymore apart from like tituration that's all I know is My GCSE physics GCSE chemistry it's just Tai history and then everything else I know I'm assuming is either biology or physics but it's probably
chemical to be honest fair enough so you're describing chemistry as just things changing color from its titration that's all I know that's all I've been told that chemistry is because I think you know orbitals and energy levels I've been taught that as physics because it's about like angular momentum and like the pulled Exclusion Principle and so I just maybe I just view it as physics um because that's the way I was taught it but I think a lot of people would say
you know energy levels is chemistry but for me I would never say that so I think it's just depending on how I think it's also a biasing I think I'm like oh biophysics is different and it still feels the chemistry is completely different so kind of going back to that whole chaos mathematician in his cool leather jacket how many times can we fit that Jurassic Park reference in I ah see I tend to think of chemistry so after my um PhD in computational chemistry I ended up doing radiation
chemistry in the lab so I tend to think of chemistry as things that involve electrons and how the electrons that orbit the atoms interact his radiation chemistry was a lot about what happens the atom when it's exposed to radiation what electrons come off it where do the electrons go do you get ionization that sort of thing which starts to become physics when you're looking at those interactions of the electrons with things so in the computational chemistry essentially relied on principles of
physics anyway so yeah they kind of all blur in my head a little bit as well rather than radiation chemistry is definitely about electrons I think one of the examples that Emma came up with when we were planning this episode was actually about simulating protein folding or at least that came up in the what's the other episode oh my gosh chaos chaos theory from my perspective that that kind of scale of atom is is that not more biology than chemistry or physics or
could you break it down to chemistry which I would understand it as functional groups interacting with each other to create a new um molecule or group I mean I think the functional groups in chemistries is what I would not to like switch it again but like in biology building up an amino acid sequence that is entirely dependent on the the functional group of the amino acid and so I feel like it I feel like it's all just um a name a name game of what what are you
going to call this area of Science and what fits with it the most if you have a physicist doing it does that make it physics maybe or maybe maybe it's maybe it's been chemistry all along um but I think I think physics is the physics part of biophysics is when you're actually like setting up the simulations using Newton's laws to be able to understand those interactions and looking at like maybe I guess the smaller scale um interactions and how you actually build up that representation and then
the biology is more how it behaves and how does that work within the body and then like I said chemistry I don't really I don't really know it's there though yeah and for me and simulations that I was doing um you weren't you couldn't really look at what the individual electrons are doing so I said chemistry is all about what the electrons do it was more about if you've already got a molecule and things are already bonded together how does that interact with other molecules or with
other parts of the molecule and that was as you say very much about looking at the physics and how these they're called non-bonded forces or non-bonded potentials how do they cause things to interact so essentially if you've got this cloud of electrons around a nucleus how does that cloud of electrons make things happen to another cloud of electrons without actually making electrons come off it and that was the limit of the computational chemistry beyond that you're getting into
weirdness things that you wouldn't do in a molecular simulation really I feel like we need to explain what molecular simulations are and how they work now so that was a very hazy example I can give that a go I mean I should be able to like I said I should be and I should be an expert on this podcast molecular Dynamics station and it does differ between what software you use of course um was to build a force field that's just like a potential it's not the same thing don't mix them up you
differentiate it's the same thing I'm confused now I heard force field and I just imagine immediately just imagined like a colorful light Dome just like appearing around you yeah even though I I know what forces are what is a force field it just describes all the forces that are there in like a in an environment oh your Force feels the forces that are there okay not a shield okay got it so yeah Laura is correct potential is is the correct word but you differentiate use calculus anyway it all
becomes the same thing um but you started off with a potential that describes uh your system and we started out with using like describing every atom and so like you said with the non-bonded interactions and the bonding interactions you're non-bonded can be your like electrostatic attraction that is happening between the atoms but also you have this thing called Van Der Val's potential and so it's basically just like a massive like equation that has all the terms that describe
interactions between the atoms and even in itself that is simplified and modeled in certain ways but then once you have this potential then you can use nuisance laws and you can differentiate to get the the force and then using Newton's Laws just like everyone knows your force is equal to mass times by acceleration and you can get kind of a data set of the positions of the velocities and the accelerations of these atoms are in your system and once you know where an atom is how fast
it's going and it's Behavior you know in the next time step then you have like a simulation and obviously you just visualize that using the data set that you have um and so that's like the app all using every atom but um when you're doing simulation but you know it's all well and good having a perfect description of a molecule with every atom being described well um but then you go to actually run a simulation and you don't have the computational power to understand how
that system behaves longer than maybe a picosecond and so you have to start having approximations and grouping the atoms together so you can have longer simulation so you can have like observe the time evolution of the system so that's my I use like and that's processes called course graining so I use coarse grain simulations to see how these proteins move throughout the membranes and how they attach um and so but even like I said like using those simulations you have
approximations and so they're not this kind of exact um way to understand how systems evolve in themselves it's like there's a lot of research that goes on to how can we make this simulation more um accurate and it's by comparing with experimental data and so there's a lot of kind of back and forth between validation with experiments and then validating that with simulations and then using that to build this up and it's like kind of a symbiotic relationship I think between them but
it's just all about the approximations in the end it's very interesting because I I never try to do a simulation on that kind of level but but um a Pico second doesn't sound very long because when I'd been in a chemist you know when I did a chemistry lab and was doing reaction like well I say doing a reaction I mean I mean experimenting you know putting different chemicals together and trying to observe a reaction um it it does feel like it happens very fast but I'm not sure it happens like
pico second fast so you know with with that kind of constraint of like single atoms being simulated to coarse grain sort of how how long can you put prolong your simulation or is that long enough to get to the um sort of reaction you want it to see or is that just simply you know you've seen the the atoms move that little bit but do they actually have their interactions yeah I think it all depends on a lot of how you set up your simulation so if you had your
simulation set up because also you can use another part of how physics comes into this because I did do a physics Masters was the thermodynamics and how you like set up the thermodynamic conditions of your simulation which is the idea is to try and mimic the experimental like temperature and pressure and so obviously if you have a higher temperature then your molecules are going to move faster and so they might reach the membrane quicker and so you might need a slower uh simulation
time in order to be basically what I wanted to observe was The Binding to the membrane and that also depended on how far did you place the peptides from the membrane to start off with if you place them too far away maybe they'll never go there um and so it was a bit of a balance between how how can I but we went for about one microsecond in the end um because that allowed like some The Binding then some Behavior after The Binding to be visualized but you know
using it at higher temperature placing them closer to the membrane could have got more long-term behavior in that one microsecond so I think you have to do a lot of like trial and errors to understand what a good time frame is but also it doesn't make sense to run simulations at picosecond light anyway because when you have like Brownian motion and like just the random motion of particles under just obviously just having some energy like kinetic energy um you don't want to interpret that
movement as being determined by some biological process when pulse particles are always going to move and so you want to try and have a bit of a long-term view on it anyway just to understand the behavior so I think it all depends on what time frame um how long you want to look for after binding before binding but also on the conditions you use anyway I think it differs a lot individual atoms and small molecules can move quite quickly so a few picosecond simulation for something
like some amines and water it's probably quite a long time yeah I think one thing one of the limitations is so you mentioned that you sort of you're solving Newton's equations of motion and using a Time step and that time step is incredibly small and the reason for that is if you use a bigger time step your atoms might have moved too far from one time step to another so you'll get this really weird unphysical result where you could end up with two atoms overlapping
where they shouldn't a new simulation will just fall apart yeah so the limit is sort of moving the atoms a really small amount for each step in the simulation and then recalculating the force each atom has on the other atoms and then using that to say well the next times that this atom will move this way so you have this picture of usually its atoms vibrating around really quickly which is what atoms do at room temperature anyway so we're all vibrating right now I forgot where a
Pico second fit on the whole like magnitude scale so I just looked it up
and it's 10 to the minus 12. a millionth of a millionth yeah yeah so million for a million for a millionth of a second oh yeah is it familiar it's a million different families does it add all times I guess 11-0 is behind the decimal point yes so very very small and in the lab obviously you're not you can't resolve that come on no no no equipment can resolve that so I guess it can also allow you to see some different behavior of different time skills if you're using simulations
but also I don't know where in the world has a computational power to simulate an experiment for 45 minutes if you're using in like an atomistic representation that would just that would drain every every memory every bit of memory that you have you normally do these simulations and I use the high performance Computing cluster and I was using sort of like 12 processes that are more powerful than the standard desktop processor at the time I was using 12 of those working
together as one different parts of the simulation running on each processor in I know computational power has become better in the intervening years it's been 10 years since I did my PhD but it's still a bit of a limitation how many atoms can you simulate in a simulation or how many how long can you run that for given that you're spreading this across different processes and you've got to consider how much RAM is required as well yeah uh has had my simulations run overnight so using 12
processes like having 12 desktop machines all working together overnight to do something and then repeating every single night and my simulations weren't that huge either they were a few thousand atoms and that's that's not massive considering how big proteins are if you think about a drop of like a single like milliliter of liquid how many atoms could be contained in that a lot given my simulation boxes are only a few nanometers on assignment yeah yeah yeah I was gonna say my
simulation boxes I think were 15 nanometer cubes so tiny very tiny simulation boxes how useful are those when when we only have a small such a such a small sample compared to the sample or is it or could we say it's the same you know that that nanometer cubed is the same as one millimeter cubed not one millimeter cubed but millimeter a milliliter one milliliter yes one sort of trick that you use in simulations is something called periodic boundary conditions where you essentially say to
the simulation that that little box I think we were both using cubic boxes it sounds like Emma but they were sort of all replicated infinitely alongside each other and you just wanted to make sure that the box is big enough so that when they were sitting side by side an atom on the left of the box that would appear on the right of another atom when it was sort of replicated wouldn't interact with the atom and with itself so as long as your box isn't too small and you can sort of
create this infinite tapestry of this um interaction then it's fine but I guess for your biomolecules that we'd want that bigger box to make sure you didn't get that sort of self-interaction effect yeah yeah sometimes I change the size of the Box because you can immediately tell like if you know you have that issue and like that's the one of the you can have like um different artifacts in your simulations from using periodic boundary conditions but in like essentially
though like over like a cell membrane not like everything averages out but it does make a lot of sense to use periodic boundary conditions and kind of break the scale down um to be able to run simulations of that size and then it's I think it is very like scientifically accurate to say well this is now representative because that's the way you build the system off anyway and so I think it's it's important to just know that when you're when you're doing both the
experiment and the simulations what what like what qualities can you compare between them and what makes sense to use one to inform the other like by looking at specific like parts of how things happen or even just comparing like if there's something happens in the simulation with one peptide quicker than the other peptide and in the lab you have the same behavior that's kind of makes sense whereas if you're comparing direct time scales that doesn't actually really make sense to do
that so I think it's just about being careful with what information do you want to extract from both in order to compare it in a way that is makes sense and also is good like a scientific practice yeah and I know for the polymers that I was looking at you talked about force fields Emma so um describing how the atoms interact using potential energy yeah there are different potentials developed for different scenarios so one potential might even really good at accurately
reproducing the density of my Polymer for example another potential I'm like you've been really good at explaining how CO2 interacts with it which is what I was looking at so you have to pick the right potentials or the right force fields for the phenomenon that you're interested in and I suppose that's probably one of the slight limitations of molecular simulations is they can't infinitely replicate all the different interactions that atoms will have in all
different scenarios so as you say it's about carefully picking what is applicable for the the picoseconds that you're looking at yeah because I think it also seems sometimes like simulations like like we said when we said oh you can just you know sit on your bed send them off it's nice and it always feels like this kind of magic solution but there's so much engineering that comes into figuring out what actually makes my simulation useful and it's in the way you set it up and
figuring out those parameters and conditions to test um you know can take a while to you know understand and optimize that and maybe arguably it could take you just as long to figure out how to set up a good simulation as it does to build a new molecule in the lab so I feel like at some point it becomes really circular where you you want to do some simulation work to stop you being in the lab all day but then you end up debugging assimilation you know just for just as
long as you would end up waiting around so I feel like it always seems a bit um of a nice solution until you actually get around to it sometimes um which is a bit interesting I definitely had that issue maybe not so much with debugging um it was just trying to find the right coefficients for a reaction I was trying to model in um not not quite in a simulation I just it was just it was just a model I suppose because it was just a sort of calculation just a bit of math yeah it
was based on on some formulas but I couldn't find the right parameter for that particular chemical reaction which it would have been way easier if I just had the materials and just did it in the lab and then timed that and then just took that into my equation but I didn't have that available um so it does sometimes feel like yeah if I could just have this in real life and maybe that's why I went towards engineering because I'm like just be easier if I could just see it [Laughter]
saying that you know with the power of computing going so far ahead you know that that curve I can't remember what's the curve of like you know computational power getting smaller and smaller do you see how how this could help say Pharmaceuticals and production of medical drugs and reducing that cost because I was reading about how expensive drug development is there is a lot of discrepancy apparently um some people are questioning if they are as expensive as the pharmaceutical
companies say it is but equally it still obviously takes a while as well to develop new medicine so how do you think computer simulations help that it can help a lot with um especially initial stages where you're just deciding what do you want to build in the lab and then test like doing the first round of testings I think if you could simulate those you could get rid of some kind of bad options almost immediately and that would really help you know money-wise if you're not
putting a lot of money I mean I guess also people's time as well um into making things I did see uh interesting reference in in the Wikipedia article on cost of drug development that they said the cheapest stage is animal testing oh yeah because it does rule out those drugs that you don't want really effectively so a lot of mice if we are able to simulate then we can avoid animal testing but in that case you if you want to know what the effect is on an entire body you'd have
to have some way of simulating that entire body or animal so you'd have to have a community simulation of a mouse and all of the chemistry that's inside it I don't think that's even possible not at the minute no even if you used all the tricks that you had to sort of coarse grain all the different parts of it so it wasn't all just atoms it was like groups of atoms but I'm not sure that's entirely necessary it depends you'd have to sort of filter out like what systems is that particular chemical
acting on that you might not have any interest in like simulating the brain at all because it's not even getting to the brain so you'd have to have some way of figuring that out I don't know if that's something that's AI could help with because you mentioned this in a recent episode about how it could help with Healthcare right that it analyzes patterns right and it uses this uh data that people have fed it to train it to figure out patterns of things that are
happening so maybe that would be one way it could help that you could say all these drugs definitely aren't going to affect the brain so we don't need to simulate that bit of it or these drugs they're definitely not going to get anywhere near your bones so we can leave that bit out this is mad speculation by the way I read that one of the biggest areas is is that drug drug Discovery drug selection in that you know someone knows I need a molecule that sort of fits this
profile you know that whole setting up the simulation also so you could say you could do the same for a drug you're saying I want I know that this particular virus interacts with this area then we need something to to have that mechanism that stops that virus and so you know trying to find the right thing that fits that particular key seems to be one of the most difficult things and maybe yeah if imagine like a machine learning script um going through different iterations to
try and create a new molecule that has all those mechanisms that it can do maybe that would be one of the ways Ai and machine learning is essentially the same thing right it's just different words for the same thing yeah I think so but with um machine learning you have to have like how good your machine learning learning algorithm is is almost directly proportional to how good your training data is because that's all it has to go off and so if we had like more extensive
training data for these different kind of um you know mechanisms in the body I feel like we could get to the point where AI becomes really good at recognizing what's going to be good and what isn't but I feel like it's so like one of the main difficulties with machine learning is having training data that works and you train your machine learning you're like yep great amazing move on to the next training set and it completely just does not work because it's got so used to the training data
and so you need to have all these different sets of training data in order to actually confirm that your um your like algorithm now works when it's been tested multiple times and so I feel like it's actually so difficult to come I found just from speaking to friends who had some projects on machine learning is the actual training data size they have is so small um it's really hard to actually build up I mean I don't know what's happening in the big research groups at different
universities but um just in like a small like little Masters project um it was actually a huge problem that they had and so I feel like it's maybe it's just there's limitations to everything anyway I wonder if that's one way that AI could help you're only talking about trying to set up your simulation initially and it does take a bit of like understanding it well enough to know which force field to pick or which potentials to pick where to put your molecules initially so you're not
creating all this really unphysical stuff there should be a massive amount of data out there from all the people that have tried it before on any simulation scenario you can think of like so I was looking at polymers you look at drug delivery and the people were looking at crystalline materials and what they do so you could take all of that information about how to set up a simulation and what worked and what didn't and feed that into the AI so it could then predict oh you want to see
this simulation so you set it up this way and it takes a lot of that leg work out that would be really good I wonder how much storage we would need to be able to contain all those especially all the failed um experiments because you know papers are generally like I did this and it has a great result very few like there might be some discussion of like I tried this and it failed but you're not going to really go through all the knits nitty-gritty unless that was part of the
novelty of your research but is true that is definitely something you can crowdsource from PhD students though so I made a lot of probably really obvious errors when I was setting some of my simulations up like I'd created black holes yeah are you really intense brushes in a nanometer box wait I thought last episode we were talking about how I was amazed that someone else created an artificial black hole and you've made a black hole I didn't say that last week essentially
yeah when the money calculations are saying oh the pressure is not a number anymore it doesn't really apply though but I've had water doing all sorts of things that water probably shouldn't do it wasn't all spread out nicely in the Box it was forming all these like weird little strings and I just I'd set up the simulation wrong that was it see and you have to add that knowledge that water does not behave like this exactly yeah so yeah get all the PhD students when they're learning this
stuff to just feed it into a database and let the AI figure it out brilliant yeah imagine how much faster it would be I feel like in the in the sort of starting phase when you're trying to set everything up you go through so much failure first if you just knew like how the setup should have been and then you could have just worked on the experiment that would have been so much faster wouldn't it so there's this thing called the protein database right which is
where people have figured out what the structure of these proteins are and you can just download the structure and go and it's pretty quick um I assume anyway I never worked on them so I couldn't say but I think a lot of the legwork in that respect has been sort of taken out and you can concentrate on doing your computational experiment because people put the effort into finding out what the structures are um and then I guess the question is what happens next so protein folding I think
is still quite difficult to figure out it's really complicated I read something that said that okay so you've defined your potentials for your protein in this instance but then as the folding continues the potential energy is change which means the potential set that you're using to do the simulation has to change and at what point do you know that it's made that change as well yeah and I imagine it it's not like a switch being flipped it's a gradual transition maybe
that's something else that AI can help out with because it can learn when that switch is occurring and change over how the simulation runs then yeah simultaneous differential equations that's always fun you just scared me then [Laughter] but isn't that something else that is essentially happening in the simulation so yeah yeah you're solving what thousands of equations of how yeah this atom has these four stacks on this atom and that atom and that atom so you're
sort of solving way too many equations to even contemplate doing by hand and there's obviously a limit to what you can do in the simulations Right In traditional molecular simulations maybe that's something else AI can help with if it's solving this sort of many body problem yeah yeah I think like a lot like in terms of computation power AI can definitely speed through you know sorting through what's actually useful to doing what's not like there's a way to train it to do that that can
just save a lot of memory in itself before you even run a simulation to check you like I don't know find a way to check if that's even going to be worthwhile running and then you can finally save not only time but a lot of memory and and money uh in that way too yeah and I read a few news articles that have said that AI is being used to develop drugs and it is definitely brought down the development time but they didn't specify exactly how which happens a bit frustrating yeah but I can
also see if you're a company it's designed an artificial intelligence tool to take the leg workout you would definitely want to keep that a closely guarded seat yeah yeah these companies have to make money company secrets and their money making ways I guess we'll never find that out unless we're there and then we can't say anything about it so I think that's a good place to leave it I've definitely learned a lot from uh Emma and Laura about how biophysics and simulations
work I see the link sort of working with both atoms and bigger scale simulations could really help with understanding how the world works so I'd say yeah that's a good end to the episode thanks for listening the views expressing this podcast belong entirely to the person that said them they do not represent any industry or organization if you enjoyed listening to these views it would really help us out if you could rate US leave a review and tell a friend this podcast was sponsored
by no one but if you're interested in funding us to continue to have Frank discussions about science and engineering please get in touch [Music]