 Okay, so it's my great pleasure to have Michael Levitier with me and he is one of the pioneers of founding molecular simulations as a field and we have him today and we're going to discuss a bit about where the field has been, how it's progressed and where it's heading today. So we can start a bit with how everything started, Mike. You can pretty much tell us, how did this journey start for you? Well Iman, it's great to be here. I love Sweden, at least it's light outside so you're worth it, that's very positive. So it started a long time ago, close to 50 years ago. It started I think, I thought it started by accident but I think it was actually directed by some very powerful people who knew where they should go and the idea was basically that once we saw protein molecules, had very detailed structures, it suddenly became why? It's not like grains of sand on the beach, we know why they're random. We had to understand the forces that stabilised proteins and that was the initial impetus. We started with minimisation, normal modes, then thermodynamics was an obvious thing and you know that is now 40 years ago that we've been doing simulations of various kinds. Simulation normally means molecular dynamics but I think simulation should include all the different techniques, Monte Carlo, normal modes, minimisation, overtaking to all parts of simulation. So it started a long time ago, in those days computers were very very weak, something like 10 to the 10 times less powerful than today and as a result we had to make things very simple. I guess we got the simplifications right because they're still being used today, which is scary. And when you started you had, I mean the application areas were limited due to the computational power you had and as you mentioned you worked with a lot of different techniques, big spectrum of techniques actually. So I think a wide range of techniques is very important because no matter where you are, you never really have enough, well if you could do a calculation just like an experiment, you'd still have to do all the analysis, there really wouldn't be any better than doing an experiment. So I think we always think about clever ways to capture the essence in terms of simplification or find methods that avoid injecting noise that you're going to take out and so on. So I think broad techniques are, it's surprisingly, a lot was very early in the business, the problems that we're sort of facing was one was how our protein folds, so this is a problem which is still with us. Another problem was how do enzymes work and again that one I think has been solved to a large extent thanks to computation. Often the computation is more quantum mechanical than molecular mechanical and then in between there's a whole range of issues but right at the very beginning it was clear we were worried about sequence giving rise to structure and structure giving rise to function. Other issues which I think were more studied by physical chemists in those days were things like pure liquids, water simulations were only done about the same time. So we actually did, I think we did a calculation on a protein before one had really been done on water which was kind of amazing but people were worried about solubilisation, free energies and things like that. So I think the problems have been with us for a long time, it's just taken a very long time to show that you could do something more useful. And what's really impressive with this field, according to me at least, is that it's not just one expertise that is needed. People have to know biophysics and computer science, chemistry, there's a mixture of different fields, physics, that made this realizable but I guess in the beginning it was also a lot of theoreticians that used this and experiments was pretty much orthogonal whereas today it's starting to merge. I think what's happened that has been very important is that these techniques have been come sufficiently alright, they're very multidisciplinary. I have a joke slide where I say we use methods of maths and computer science with forces from physics applied to molecules in chemistry, to biology and medicine. I probably left something out good for the economy and the rest of the world or something but basically you could hit every single thing and it's sort of true. I personally like that because I have very wide interests in this way, you know and you do many different things, maybe not an expert in any of them but I think greatly today these methods are very very widespread, they've become really accepted, it was a long time when computing was the ugly little sister of biology that's no longer the case and because they become widespread they are used more broadly and often by people who are not experts or experienced at a particular technique. Yeah so I mean in life science it's probably gradually becoming more biologist or I mean the application expert that is starting to look at this because they know the biological system, they have the questions and they want to apply it with one of these techniques and in your perspective what are the biggest achievements so far and what do you see? Well I'm just to relate to that at first point I think that if you think about a particular problem, for example now they're trying to work on the ribosome so calculating calculations on the ribosome is hard but its trivial compared to reading the literature on the ribosome which is incredibly hard, there are probably a hundred thousand papers you know people who have been doing this for all their lives really know it so what one really needs to do is either team up with what I call a domain expert, somebody who's doing experimentals, who else have the experimentalist do the calculation and the calculations don't really care what items they're working on, they don't care what system they're working on so the programs are much less domain specific than the experimental knowledges but I think it's definitely clear that any of these problems need to be linked to experimentalists who have a lot of experience in knowing where the hard questions are. I also think for the computer person it's a lot of fun to have this interaction so again I was lucky that all my life I was never in a purely theoretical department, I was always in a department of people solving problems and therefore someone would say is DNA stiff or flexible? It's a hard thing to measure but a calculation maybe can be useful so I think that is very very important to have this interaction I think today what has happened is that more and more as the programs become more widespread, easier to use, more and more the actual biologist is taking the program and using it and I think it's also very very good. Yeah that's very very I mean good for good for the field of molecular simulations you could say because it's no longer a niche area it's starting to spread and more people are trying to use it but as the complexity of the problems increase you need more types of different experts and also you need more computational power and there are a lot of different initiatives that try to make this applications accessible and share the knowledge and make sure that people can use them for applications of a relevant problem scope and yeah so forth and these are probably being done in each and every continent and what do you think there are? Well I think the applications I mean I would say the applications are increasingly bug free in other words not only that they probably also check the data so it's harder to do really stupid things one would hope without knowing about it so a lot of the expertise would need to sort of navigate through buggy codes I'm pretty sure that all the codes I know pretty much bug free the parameter sets if there's a missing parameter most programs will say missing parameter and not use the long parameter stuff like this I think this is being very important computers are such that you can now run quite large certificate calculations I think the hard part is when you want to do calculations that are sort of off the beaten path and I still believe that there's a lot of different ways of doing these calculations it's not simply a question of starting with a crystal structure and running it for a certain amount of compute time and analyzing the results and other things you can do those things require more expertise so I think there's definitely still room for certainly on the harder problems for collaboration between a computer person or a simulationist and an experimentalist in other cases I think experimentalists can actually do very valuable calculations themselves and you know one thing that's interesting about experiments especially in biology they're one of the really big achievements of the field is that they're very very good at worrying about controls and statistics because if you want to ask whether a certain cell line is toxic or not you've got to be very very careful so biologists by their nature are really good at looking at very messy systems and sometimes these computer systems are very very messy so a good experiment would say okay you got this calculation now let's see what happens if and they will do a control and that could be very very helpful I know this is something which should have been very obvious but it was only sort of in my mid-career when we were doing a certain test on seeing if it's full proteins and the results looked interesting but it was hard to tell whether they were significant and then I think my my student at the time had the idea to just randomize the sequence and as soon as you randomize the sequence all the signal went around we were thrilled but I mean we never thought it was necessary because we knew it was there but still when we saw that the signal did depend on having the correct sequence we were very excited so there's these kinds of tests are very very important and I think experimental biologists are really really good at doing this so I think one area that we can learn from them is how you do clean experiments on very messy systems and as you mentioned you often try to do things off the beaten path and in science you pretty much that's more the rule that you try to do something that is off the beaten path and start you know want to address larger problem areas that nobody still explore it's a personality issue I think you know in if you look at people there are people who walk on the beaten path and people who like to walk the beaten path and I think it's just my personality I don't know whether it's a good thing or a bad thing I know that earlier on people would say my grants are very unfocused and off the beaten path you know there's lack of focus I mean I think you need both things it's just that I'm very happy to do things where I don't know where I'm going and being off the beaten path is by definition there's no path but I think equally well there are there are other areas where you need both things I'm not saying one is better but I think personally you know I've enjoyed that breath and it's probably even getting wider than it used to be it's pretty wide so and as you mentioned also there is a lot of other fields such as like experimentalists that have a wide knowledge of how to validate something or do a do a type of anti thesis checking and you know it's difficult for example if you're sitting in a theoretical department who should you contact to get these types of you know you need to establish a collaboration to get this multidisciplinary I think also there's there's it's necessary to establish trust um you know a typically experimentalist approached by random theoreticians oh god this is this I don't even understand what this guy's talking about he's talking about correlation functions and this and this and this I grew up and I did my phd in an experimental lab I was always in experimental labs right now at stanford I'm the only theoretician in a lab of 10 faculty so at least that way they trust me I mean maybe they don't trust me but they know me and what works both ways I actually have the two projects I'm working on right now one involves a collaboration on orally polymerase with Roger Kornberg and he knows that system backwards the other is a much more difficult problem involves a collaboration with another faculty member Jolene Puglisi on the ribosome now he is he just knows the ribosome really really well and one of the problems is it's when you have a very complicated simulation how do you bring the experimentalist into it and we've actually got very involved in in virtual reality so we have a couple of Oculus rifts and the idea was we haven't really succeeded very much but you could make you walk into the system and the great thing would be is that you know the domain expert next he would say gee that that you know nuclear tide that's moving a lot that's really interesting which one is it and then you know he looked it and said oh wow at a meeting I went to three weeks ago somebody said that's a very important thing for this function so this is something you'd never be able to find by any kind of research are you able to sort of but this project is you know we we sort of got it working but well all that virtual reality stuff is still a little bit early hasn't really prime time yet yeah and I guess actually in you know many labs these days they're very good at combining you know theory with experimental but there are many more that probably wants to do this transition and I mean for example in this building there is a wide variety of experimentalists and theoreticians but if you look at this at a larger scale on a continental scale or global scale where you want to make these connections and I don't know if these big initiatives help I mean the researchers are very good at collaborating still but is there anything that can you know make the push so you can make a bigger leap there so I think there are two ways of going and a lot of of the young people today maybe not so much in biological simulation but say in bioinformatics are doing experiments and calculations themselves I think that's good I think that probably only works when the experiments which often just involve a sequencing machine are relatively easy I still think that collaboration the great thing about collaboration is that it's a meeting of two minds if I meet somebody who's been working on say a membrane system all his life his brain is just teeming with this and I can interact with him and that is really it gets really really exciting so I I still believe we know we don't want everyone to become a mixed theoretician experiment this we still want specialization I think collaboration between people who have specializations and are different always leads to more than just having two people who are both doing the same things so I think that's still gonna be important I think collaborations you know one of the really nice things today is that because of the internet and and it's very accessible it's very very accessible that this morning my wife phoned me up on FaceTime without without the picture and the sound quality is better than any international call I think I've had in the last five years she was eating her breakfast and said wow you know I can actually taste it and she called me on FaceTime and said there's an answer you know there's all you have the channels I'm just saying these things can be very very powerful I think that the key thing is to have a wide net the reason you want to have an international kind of center is that you're able to capture a broader sense of people as I said the key thing here is collaboration between people who are not the same they've got to have a nothing common that they have a common but even I would say that they would actually be on a high scale biology and then for example but it has to be quite different if you have people who are very similar then it maybe seems easier it goes well at the beginning but the potential gain is less so I think you want to be able to have you know a broad capture area and then of course bring in experimentalists as needed I think it can be very very powerful and the last thought we would like you to share with us what do you think you know from today and onwards what are the greatest challenges that we can you know try to address and hopefully solve and what timescale do you think we can solve them I mean using both computational and experimental techniques together so I would say there's a general challenge which we make progress on that needs to be solved in a more specific one so I would say the general challenge is that it's quite easy to generate a simulation we can generate big data any size you want but actually analyzing to making a movie is is always easier and not not trivial but a good movie is a hard thing to make but it's one way but it doesn't it isn't enough we need to find various ways of analysis analyzing these things as I said virtual reality might be used for different ways of analyzing numerically different projections of the space etc I think that's something which is going to get better I think there needs to be maybe more emphasis placed on improved analysis methods I think the for me the big challenges are systems where they're so big that for example a ribosome and a box of water is something like more than a million particles maybe two million particles and what makes it so hard is that firstly there's just a lot of data but secondly there's a lot of noise so you need to understand how you reduce degrees of freedom and I think one of the big challenges in any mechanistic study we don't have to find motion we want to find functionally important motion and you know most biologists think in terms of a reaction coordinate you know go from product reacting to product if one dimensional so one dimension might be too simple we need to get very complicated systems actually have a two dimensional picture but we're taking a space which might be six million dimensional like ribosome and projecting it down to two and we don't know yet how to do that very very well so I think there's real challenges in choosing the right reduced sets of degrees of freedom and that's a very general challenge so I think that also on these very large systems we have to find a way to reduce the noise we you know you can't run a full ribosome simulation for enough computer time to get rid of all the noise so we have to try and do the calculation in ways that don't generate the noise in the first case and these are techniques which I don't know we have the answer for but I think are interesting and I think that these machines are really interesting because they have moving parts and and it's not just the movement if it's productive movement when you think about a molecule vibrating it's vibrating because it has no choice when a ribosome does what it's doing it's part of the action so it's very interesting well thank you very much Michael and thank you for sharing your thoughts with us and we hope you have a great continued stay here in Sweden thank you very much your money it was really nice talking to you thank you