Computer Modelers Secure Chemistry Nobels

← Back to Stories (view on slashdot.org)

Computer Modelers Secure Chemistry Nobels

Posted by Unknown on Wednesday October 9, 2013 @04:40AM from the cyberspace-like-space-but-cyber dept.

ananyo writes "One day, computers may be able to simulate exactly how enzymes, ion channels, viruses, DNA and other complex biological molecules react with each other inside a cell. And if such a software package is ever written, it will owe its development to three researchers who today won the Nobel Prize in Chemistry: Martin Karplus, of Harvard University and the University of Strasbourg, Michael Levitt, of Stanford University, and Arieh Warshel, of the University of Southern California in Los Angeles. Starting in the 1970s — working with computers far less powerful than today's smartphones — the three theorists made advances in computer modeling that laid the foundations for modern software used to simulate protein folding, design drugs and even artificial enzymes, and understand the workings of complex catalysts. In essence, says Sven Lidin, the chairman of the Nobel committee, they 'took the chemical experiment to cyberspace.'"

34 comments

Min score:

Reason:

Sort:

Heisenberg by Anonymous Coward · 2013-10-09 04:51 · Score: 1

Say my name.
Re:Yawn... by Anonymous Coward · 2013-10-09 05:10 · Score: 1

Having one and EARNING one are apparently very different these days.
If I was Al Gore, I'd be pissed.. Really!
vindication by Anonymous Coward · 2013-10-09 05:12 · Score: 0

This reminds me of my three year long failed project in computational drug design. Sigh ..
1. Re:vindication by Anonymous Coward · 2013-10-09 05:27 · Score: 1, Insightful
  
  You should try Climate Modelling. You can fail there and no one cares. They will even endorse your results regardless of their accuracy!
2. Re:vindication by Silvrmane · 2013-10-09 05:35 · Score: 0
  
  You say this on the day when I don't have my mod points. Otherwise, I would have given you an Insightful.
  
  --
  planet texture maps and more
3. Re:vindication by plopez · 2013-10-09 06:06 · Score: 3, Insightful
  
  I did my project in groundwater modeling. Modeling is a black art.
  
  --
  putting the 'B' in LGBTQ+
4. Re:vindication by Anonymous Coward · 2013-10-09 06:09 · Score: 0
  
  Not to worry, nobody else has ever managed to design a drug using computers either. A lot get published in the field, sure, but that's just for "academic curiosity", little of which leads to any actual advances or insights. As for classical molecular dynamics, nobody even knows how accurate the parameters (ie. force fields) are and different, "equally valid" ones give different results. A guy I know published over 20 papers during his 3 year PhD in the same exact field that the Nobel prize is for. If you can get that many publications in some field, that field will quickly turn to be self-sufficient as the researchers will be citing each others' work without much regard for what is being done by others (say, by experimentalists, or people using different techniques).
  Not to say that seeing molecules on your computer screen isn't cool and all, but it isn't as useful as one might think.
5. Re:vindication by riverat1 · 2013-10-09 14:29 · Score: 0
  
  Or you can do better than any other method we currently have and still get called a failure by bozos who don't have a clue of how to evaluate the results in the first place.
6. Re:vindication by Anonymous Coward · 2013-10-09 16:28 · Score: 0
  
  You cannot begin to imagine how happy this makes me as a chemist and computer scientist I'm thrilled that my first passion, computer science, is being recognized by a Nobel in my second passion.
Just wondering... by TheloniousToady · 2013-10-09 05:29 · Score: 5, Interesting

Has a Nobel Prize ever been awarded before for an achievement that was specifically software-based?
1. Re:Just wondering... by Anonymous Coward · 2013-10-09 05:34 · Score: 2, Interesting
  
  i don't know, but I have met Martin at a few meetings over the years and his code(s) (CHARMM etc) really did change the way things were done. Bear in mind the classical physics hasn't changed in 150 years but the quantum mechanics was a relatively new 50 years old!!!
  And being software, ANYONE can use it....!
2. Re:Just wondering... by ibmleninpro · 2013-10-09 05:44 · Score: 5, Informative
  
  Kohn & Pople's prize in 1998 is probably the closest in relation. Pople's work on ab initio methodology is directly related to the GAUSSIAN quantum chemistry suite, and when he become unassociated with GAUSSIAN due to disputes (and was subsequently "banned" from using GAUSSIAN ever again, see this site for criticism regarding this -- note that for some reason Chrome is saying this is a dangerous site though I've been on it many times, so click at your own discretion -- or just google it) he went and started the Q-Chem program suite.
  
  I don't think Kohn was ever associated with any quantum chemistry program suite like Pople, but he was one of the key players in the development of density functional theory, which is available in pretty much every quantum chemistry suite and used by a vast majority of chemists who use calculations in their work.
3. Re:Just wondering... by Anonymous Coward · 2013-10-09 05:48 · Score: 0
  
  John Pople?
4. Re:Just wondering... by Anonymous Coward · 2013-10-09 07:17 · Score: 0
  
  Kohn & Pople's prize in 1998 is probably the closest in relation. Pople's work on ab initio methodology is directly related to the GAUSSIAN quantum chemistry suite, and when he become unassociated with GAUSSIAN due to disputes (and was subsequently "banned" from using GAUSSIAN ever again, see this site for criticism regarding this -- note that for some reason Chrome is saying this is a dangerous site though I've been on it many times, so click at your own discretion -- or just google it) he went and started the Q-Chem program suite.
  I don't think Kohn was ever associated with any quantum chemistry program suite like Pople, but he was one of the key players in the development of density functional theory, which is available in pretty much every quantum chemistry suite and used by a vast majority of chemists who use calculations in their work.
  Mod up parent - excellent info!
5. Re:Just wondering... by Darinbob · 2013-10-09 11:16 · Score: 1
  
  Well not since I hacked their computers to grant the peace prize to Obama.
Yeah right by plopez · 2013-10-09 06:17 · Score: 4, Interesting

"One day, computers may be able to simulate exactly how enzymes, ion channels, viruses, DNA and other complex biological molecules react with each other inside a cell. "
An exact solution to a system of non-linear equations with no closed form solutions. You can optimize the functions but there is no way, except empirically, to verify them. And then your solution may be very fragile, if boundary values change your solution may no longer work. The are not even solutions but approximations.
BTW, I did a project in Groundwater Modeling. Chemical Modeling is even more complex.

--
putting the 'B' in LGBTQ+
1. Re:Yeah right by MichaelSluydts · 2013-10-09 06:36 · Score: 3, Interesting
  
  While true, it should be stated that the problems are a bit different. In quantum chemistry we are trying to solve the Schrödinger equation which we know, the problem lies in describing certain terms (many electron interactions) and then the approximations made to make the calculation time still reasonable. The bigger the system, the more compromises must be made, but there is an idea of what information we lose: upper bounds, lower bounds etc.
  If you really had an infinitely powerful computer, in a way using complex theories such as quantum mechanics makes things simpler because you are dealing with the real fundamental interactions, where the true solution must come from, whereas in less heavy simplified models we try to take into account this behavior with simplified models which in the end lose information while still having to describe the same behavior.
2. Re:Yeah right by Anonymous Coward · 2013-10-09 06:45 · Score: 0
  
  You have no idea what you're talking about and clearly are completely ignorant of the kind of models the field is using. Chemists are not afraid of solving non-linear equations, not even a system with millions of them coupled, that's routine. Most of what they do is more non-linear than the movie Primer. The problem comes with system size, as the algorithms fundamentally scale very badly with number of particles. That's why you have to have different sets of equations for different scales and connecting these to have one unified, self-consistent picture is one of the current major paradigms in the field.
  The kind of modeling you as an engineer are used to is something completely different from simulations, which are essentially experiments inside a computer. I'm not saying that the equations are different (which they are, and arguably the ones in quantum chemistry are more difficult to solve). What I am saying is that the whole paradigm is different.
3. Re:Yeah right by Anonymous Coward · 2013-10-09 07:13 · Score: 0
  
  You are not solving Schrodinger's equation if you are going the classical route, though. In which case plopez's response is quite a bit more accurate than your statement.
4. Re:Yeah right by Anonymous Coward · 2013-10-09 08:43 · Score: 0
  
  but ab initio methods are often used to provide parameters, and since some of Karplus's work was directly cited for this innovation, this is why it is so important.
5. Re:Yeah right by plopez · 2013-10-09 19:12 · Score: 1
  
  Now, how are these parameters tested. BTW please forgive me I am a modelier but not a chemist. But I understand Maths
  
  --
  putting the 'B' in LGBTQ+
6. Re:Yeah right by plopez · 2013-10-09 19:20 · Score: 1
  
  Well I am not an engineer
  Yeah non-linear models "solved" that been there.
  I understand the paradigm is different but we are still in a realm where human judgement can have a huge impact on a model. I will not say that I am a better chemical modeler than you you are, but what I will say is that we need to understand our models, the limitations of them, The limitations of the Maths behind them,
  and how difficult it is to understand our problem domains.
  
  --
  putting the 'B' in LGBTQ+
Too late Nobel committee by Anonymous Coward · 2013-10-09 07:17 · Score: 0

You threw away all credibility when you gave one to Obama. Now the award is a joke.
1. Re:Too late Nobel committee by methano · 2013-10-09 09:00 · Score: 4, Informative
  
  The peace prize is actually awarded by a Norwegian group and the science prizes by Swedes. So you can keep your respect for the science prizes and still cling to your crappy politics. Lucky you.
2. Re:Too late Nobel committee by SleazyRidr · 2013-10-09 09:53 · Score: 4, Insightful
  
  They threw their credibility away when they gave it to Obama? They still had some left after giving it to Kissinger, or Arafat?
  
  --
  Is 1563649 a prime number?
3. Re:Too late Nobel committee by Anonymous Coward · 2013-10-09 10:16 · Score: 0
  
  They threw their credibility away when they gave it to Obama? They still had some left after giving it to Kissinger, or Arafat?
  Those later two had been political heavyweights for some time before they received their prizes, most of the criticism is that Obama was pretty much a nobody on the world stage when he received his.
4. Re:Too late Nobel committee by riverat1 · 2013-10-09 14:43 · Score: 1
  
  And you threw your credibility away when you conflated the group that awards the science Nobel Prizes (the Royal Swedish Academy of Sciences for the physics and chemistry prizes) with the group that awards the Peace Prize (the Norwegian Nobel Committee).
5. Re:Too late Nobel committee by SleazyRidr · 2013-10-10 02:51 · Score: 1
  
  So, giving it to people with a history of human rights abuses is better than giving it to someone who at least said he wanted to stop the abuses?
  
  --
  Is 1563649 a prime number?
Mod parent troll by OneAhead · 2013-10-09 15:23 · Score: 1

Are you the same person who posted this? I can assure you you're just as wrong as you were back then. Since you're acting like a broken record, allow be to return the favor:
Raltegravir (trade name: Isentress) is one example of a molecule that wouldn't probably exist without classical computational chemistry.
[You are] living 10 years in the past. New examples [of computer-designed drug in clinical trials] are popping up on a yearly basis at medicinal chemistry conferences. I've personally witnessed a lead structure coming out of a modeling study that only needed a little experimental validation and optimization to give rise to a drug candidate that is now in clinical trials.
Computer modeling has been badly oversold during the 1980s, resulting in a period of disillusionment in the 1990s. Some people (mostly at non-US academic institution, I'm sad to say) apparently never got over it. To those people: wake up, it's 2013! Computational techniques have become better, and computer power has increased a few orders of magnitude. Equally important, computational medicinal chemists have come to understand there's no free lunch: a simple calculation will yield simplistic results. To get truly predictive results, a labor- and computer-intensive project involving an diverse palette of computational techniques is required, and collaboration with experimentalists is a must. Conversely, any experimentalist who doesn't have a modeling expert among his/her collaborators by now deserves the imminent outsourcing of his/her job to India or China.
To continue replying on your present post: there do exist rigorous ways to benchmark the accuracy of force field parameters against experimental observables; it's just that only a handful of experts routinely do this. And that's exactly the problem with this field - it has a bad name because of all the people doing crap, and patting each other on the back in true circle-jerk fashion (so you're right on that account). Like your friend; to do things right in this field is more difficult than Quantum Chemistry, so unless he is a supergenius, I cannot imagine 20 papers written in 3 years by a junior grad student to be worth a dime. Indeed, I'm reading papers ranging from utterly worthless to deeply flawed on a daily basis. But that doesn't imply that the whole field is flawed; just that a large number of its participants have no clue what they're doing. There's a small elite of experts who do, and they're almost consistently getting good results.
"Conflict of interest" notice: one of the 3 nobelists in TFA is my mentor's mentor.
1. Re:Mod parent troll by Anonymous Coward · 2013-10-09 20:04 · Score: 0
  
  I'm not the same Anonymous Coward from the other thread months ago. But to some extent I do agree with him.
  So, Raltegravir. This is a drug FDA approved 2007, so you are basically saying that 10 years ago they had the computational resources and models accurate enough to design this drug? I'm not saying there hasn't been computational work on the drug after it was actually designed, this is true for pretty much all drugs, as it's an easy way to publish rather than to perish. Now, in the case of Raltegravir, it being the first of its kind of drug, I'm not even going to say that the computations provided no crucial new insights, they probably did give some ideas to researchers working on new drugs as to what kind of experiments to do (like computational work prior to Raltegravir might have contributed some qualitative hints to its development). But to say that Raltegravir wouldn't exist without classical MD is an overstatement.
  As for the guy who wrote over 20 papers during his PhD, some of them made the top journals of chemistry (if one ranks them by impact factor; not to go into debate whether this is a good way to rank science or not), and he is associated with the development of CHARMM, the force field which the Nobel prize was in part awarded for. Not to say that the guy isn't smart, but my conclusion is largely similar to yours; that this is a symptom of circle-jerking; and also a cause, as if it is too easy to publish papers (structure from PDB, simulate, publish, add salt, simulate, publish, rinse and repeat), that is what people do, instead of focusing on new science.
  Computer simulation is a new and exciting field that offers... pretty pictures. Let me explain. Thanks to CHARMM (and other developments), working in the field is so easy that anybody can run simulations if they have access to a small cluster. If you don't have a lot of time, it is also a lot easier to get a qualitative picture of what is going on by looking at MD simulations instead of a quantitative statistical mechanical derivations, which typically are mathematically quite challenging. The simulations offer pretty pictures, mathematical theories do not. The fact that simulations are so easy to do and to understand (qualitatively) leads to what you said: not checking if certain parameters fit, and to the rediscovery of theories sometimes centuries old. For example I've seen the Oseen tensor rediscovered in a publication that appeared in one of the top journals. Not that they rediscovered the mathematical form, but the flow pattern, which they in no way managed to connect to the theories of hydrodynamics. Now, that the authors did not recognize the connection could be explained by the fact that they might have been biologists, and maybe it is unreasonable to expect biologists to know about the theories of physics. What is deeply worrying to me, though, is that this got through peer-review and through people who should know better. This is circle-jerk at work, and it has a feed-back mechanism that keeps devaluing the field: the peers selecting who to promote to academic positions or whose manuscript to accept to a given journal themselves produce bad science and often don't recognize the gems from all the garbage. I'd like to go on rambling about the feasibility of peer-review in the 21st century and draw parallels to how businesses have already moved (or started to move) away from structures assuming that the well-established are the well-informed. I'll refrain for now, as it's not really on topic anymore.
2. Re:Mod parent troll by OneAhead · 2013-10-10 10:32 · Score: 1
  
  Yes, Raltegravir. Not a neutral source? Fine, here's another one: Zanamivir.
  About MD being easy, I beg to differ. The only thing that's easy about it is to do crap. Computers are garbage in, garbage out. It is possible to do things right, only it's much, much more difficult. In fact, so few people know how to do things right that the crap easily passes peer review. It's not so much that the reviewers let the crap pass on purpose in a "you scratch my back, I scratch yours" fashion, it's that a lot of the reviewers simply don't know any better. To put it differently, there are far too few reviewers who know what they're doing to reliably block crap from getting published. And this is exacerbated by the attitude of those who do crap. Got a bad review? Reviewer had a bad day and is being unreasonable - submit same stuff to different journal. There are enough journals, and the chance to get a knowledgeable reviewer who can recognize the flaws 2 or 3 times in a row gets almost vanishingly small.
  The problem is not so much the computational resources or accuracy of the models. Although these have improved dramatically, even 10 years ago, highly significant results could be gotten out of them, assuming very careful setup, analysis and interpretation. That is what a lot (but not all!) of computational studies are lacking.
  To end on a more positive note, post-publication, the really good stuff does get recognized by the community and often gets highly cited. As you alluded, evaluating scientists in the field by number of citations instead of number of publications and impact factor would go a long way in rectifying this crooked situation.
Re:Yawn... by Anonymous Coward · 2013-10-09 19:29 · Score: 0

I feel urged to quote the thoughts of a bowl of petunias.