*Exactly*. And Intel's C/Fortran compiler is F A S T. (I use it on Itaniums as well. Not that Itaniums are all that fast, but who cares if you have 100+ of them.;) )
For Linux, the Fortran compiler is also free right now (as in beer, for non-commercial use), which allows me to code and test from home on my x86 and then recompile it on Itanium. Easy!
Java performance depends entirely on the JVM. If the JVM sucks, then the Java program runs slow. The JVM may be good on PIV x86, but what about Power series? What about Itanium? Clusters don't often use x86's because they consume lots of power and require lots of cooling. And new computing architectures require... a new JVM! So when the Cell processors are available for scientific computing, C and Fortran compilers will be written to take advantage of their amazing capabilities. How quickly do you think a JVM will be available??
I read the website you cited and lot of the reasons for its 'Java is faster than C!' aren't true. Most of the reasons why Run-time compilation are complete BS. A good Fortran/C compiler will do the _exact same thing_. The website doesn't convince me. It has too few benchmarks and does not provide the code it uses. Not impressed.
So, for scientific computing, Java is still not the answer. And, I'm not alone. If Java was faster, then don't you think IBM/Intel/SGI/HP would create nice JVMs on all of their HPC systems with nice development environments. They don't.
I completely agree, but you'd be surprised how many scientists (computer scientists even!) will complain that XYZ algorithm is too computationally intensive...and then you find out they implemented it in Java.
It's like "DUH!" C or Fortran. Pick one. Or C++, even. But Not Java.
Of course, when you broach the topic, you will hear things like 'C is old. Fortran is older!' Wtf cares? It's F A S T.
It's an MPI environment. MPI stands for Message Passing Interface. It's a library of subroutines that allow you to send data from one processor to another. So if you wanted to process a 2 gig text processing, you could divide the text file into 500 chunks and have each processor perform some function on its chunk. Then you would collect all of the results together and save them.
Certain programs are more 'parallelizable' than others. The programs I run (and code) are very 'embarrassingly parallel'. That means that there is very little communication between each processor while the program is running. Certain programs require a large amount of communication between processors and a lot of special attention must be given to optimizing the architecture of the parallel program.
The bandwidth and latency between processors is pretty good, but no where near the actual speed of computing. So you need to balance the time it takes to send data from one processor to another and the time it takes to compute some results. Since the processor can do both at the same time (send/receive data while computing), it becomes a sort of art to synchronize it all together so that no processor is left idling with nothing to do.
I was glossing over the differences (for the Slashdot crowd) between microRNA's, snRNAs, RNA interference, and other small RNA transcripts that bind to mRNA and perform X function.
I hadn't heard about RNA's regulating chromatin compaction. In yeast, is the mechanism known? Does it help catalyze acetylation? Does it inhibit histone binding to DNA?
It'll more likely be translational control via RNA.
RNA can quickly hybridize with regulatory regions of mRNA and change their translation rate.
And these RNA transcripts can be very small, but still regulate the translation of many genes. It'll be a while until the function of all of these RNA's are understood.
NetCDF is used more by non-computer scientists who need to store lots of data. It has an easier-to-use API. But, HDF5 can do more useful things. When NetCDF uses HDF5 as its underlying format, it'll get the best of both worlds: good API, good data structure.
When I started a software project about two years ago, I looked at both NetCDF and HDF5 for data formats. I chose NetCDF and have had zero problems (and it's been very easy to the software working nicely). I think using HDF would have added another 6 months of development time.
In the end, it won't matter: they'll both be equivalent.
I agree with a lot of your comments. There is a reason why online peer review is not going to happen for a long time.
But, for some fields, the first steps towards online peer review might be to open up commenting to online readers. Maybe the journal requires a sign in process. Maybe the journal allows anonymous commenting. Who knows. Clearly, there are problems to work out.
Also, if the field of research is small then the only problem is to make sure each interested professor is connected to the online peer review process. Social networking really isn't the problem: all you need is a invite/nominate system that encourages professors in similar fields to review each others' papers.
But, I think, a 'field' of research is not a concrete boundary. Scientific papers should be understandable by any professor in a similar field. I've often reviewed papers or grants that have very little to with my own research. That's fine...you just have to read into the subject a bit more. Likewise, my papers have been reviewed by people who have no clue what I am doing.;) Usually, they spend the time and come back with intelligent comments. Sometimes, not...but that can always happen either way.
Well, I think the process will eventually evolve to take more advantage of the internet. For example, the PLOS journals and others are all freely available under the Creative Commons License. The review processes for many journals all take place over the internet and not through mail.
How long will it take before a journal turns on online 'comments' below each article? Maybe you have to log in to reply, but it's a start. The next step would be to allow randomly selected members of an online journal to read a paper and review it. The step after that might be to designate a portion of the journal as "pre-published" and allow anyone to review it with only well-reviewed papers allowed to transfer to the "published" portion.
Of course, I'm just guessing. Credibility is the only asset a journal has. If they modify their review process that credibility may be damaged. Who knows..it might take new upstart journals to revolutionize the review process.
Though, imagine the Slashdot community trying to review a technical paper. Do you think that's a good idea??;)
The paper stating that ~50% of scientific papers are false is published in the Public Library of Science (PLOS) Medicine. The paper only examined medical studies and not scientific papers on physics, chemistry, engineering science, (and mathematics).
While molecular biology papers can be prone to statistically insignicant, but factually stated conclusions, the biggest culprits are clinical studies and 'large-scale' analyses of data.
Good experiments are constructed to give a 'yes' or 'no' answer based on the presence or absence of evidence. The zeal of high-throughput studies and analysis have put more pressure on good statistical analysis. Unfortunately, statistical analysis requires math...which sometimes eludes doctors and biologists. Hence, the problem of missuing statistics and stating inadequately supported conclusions.
I hate people who say "It is a pity that... " when they are flat out wrong. If you're going to be so condescending to someone, you better make sure you're absolutely correct. Otherwise, not only are you an idiot, but you're an arrogant idiot.
Anyways, wtf says "It is a pity that... " anyways? Do you even know what the word 'pity' means??
But looking in his JACS paper I see what you mean, he's doing a dynamics simulation.
Ah, gotcha. That does make more sense.
Right. But the inaccuracy of DFT methods is about 3-7 kcal/mol
I'm not sure what method was used, but I've seen errors of less than a kcal. You need the error to be that low to simulate biochemical reactions (which often have activation energies of only a few kcal). It might have been York's stuff, I forget. ('course, this isn't really my field.)
It requires a full potential energy surface to work with. And that would be either empirical (MM, which is incapable of breaking/forming bonds) or non-empirical (QM, which is way too expensive to do any kind of PES scanning).
I mentioned Darrin York because he recently gave a presentation on semi-empirical QM potentials. Basically, he performs hundreds of ab initio simulations of a certain type of reaction (he chose phosphorylations by a kinase protein) and then fits a semi-empirical model of the force field of the reaction. The potential is very accurate, but only for phosphorylation reactions. Then he performs QM/MM simulations using the semi-empirical force field, which is much faster now that the original ab initio calculations have already been completed.
That's wrong though. The transition-state is easily defined That assumes there's only one transition state. Really, there's an ensemble of transition states where the potential energy is at a local maxima. The potential field in the large 6N-dim phase space is never very smooth, anyways, so there's always a ton of saddle nodes around the transition state 'region'.
Then again, there could be more than one 'region' of the phase space with local maxima. These are simply alternative mechanisms of the same reaction.
But I think you already know this and were simplifying it for me.;)
My advisor does MD simulations of protein-protein and protein-DNA interactions as well as kinetic Monte Carlo (mesoscale stochastic simulation) of large networks of biological components. (The latter is what I do.)
They're not revolutionizing the methods, per say. They're using them in different ways. For example, a group at MIT created 'BioBricks'. They are segments of pre-synthesized DNA sequences which can be ligated in any order you wish using the same exact procedure. The bricks are specially constructed (with specific restriction sites) so that the same protocol can be used over and over again to continually ligate together more and more BioBricks.
That way, you identify what you want to build and then you ligate together the bricks that contain each individual 'part'. They're 'replaceable parts'.
But they use the same ol' genetic engineering techniques to splice, mutate, and ligate those bricks: restriction sites, PCR, electrophoresis, antibiotic selection.
The methods that you listed are usually used for mammalian cells. Most of these guys use yeast or bacteria for their work. If you're working on bacteria, you can create single or multiple copy plasmids or integrate a plasmid into the bacteria's chromosome. If you're working on yeast, you can use homologous recombination to target the location of integration.
For bacterial and yeast cells, it's very well known how a gene is transcribed and translated into protein and what other proteins are involved. Because we know how the system generally works, we can accurately describe it using mathematics. Then we can use the mathematics to predict what will happen if we cut out portions of the system and paste them in different combinations.
What I was talking about is the prediction of the what and not the how. For instance, given a DNA sequence for an enzyme, you cannot predict its structure. (yet)
Ok, yes, the protein folding problem (for any protein) is not solved yet. However, using transition path sampling, it's possible to fold a small 40 amino acid protein from a random structure. Google Folding@Home for details.
And even given the structure, you cannot predict the function an enzyme. (ab initio, anyway. Homology comparisons can help do it of course. But since we're talking 'deep understanding' here, I take it to mean ab initio)
Ab initio calculations are limited to ~200 atoms. Yes,... atoms. The current solution to that problem is to use a hybrid Quantum Mechanic / Molecular Mechanic (QM/MM) method that only uses ab initio calculations for the atoms located near the catalytic site and use regular molecular dynamic simulations for all other atoms. You need to merge the two together self-consistently (not easy), but people are working on it and making good progress.
So, you're right when you say "you cannot make an ab-initio model", but if the end goal is to _predict_ the dynamics of a whole cell then one doesn't need to perform quantum calculations _on everything_. And the calculations don't need to be performed at the same time.
For example, you can take a single DNA sequence and run a simulation to predict its native structure. You can use homology (on both the sequence and structure) to estimate where the catalytic site is and what type of substrate likes to bind. Then you perform another simulation to predict the reaction mechanisms, kinetics, etc of that enzyme.
Then you repeat that process for every protein in a single cell. (Leave, grab a cup of coffee, spawn four generation of children, come back, get the results.;) )
Then you take all of that data and use a different type of simulation method to predict what will happen when all of those proteins are thrown together. You don't need to use ab initio to do that (or molecular dynamics simulations).
Now, of course, there's experimental data out there so we don't need to in silico predict _everything_. We can jump start the process of predicting the dynamics of biological systems by only using well-characterized proteins and genes. And the mesoscale-type simulations of those systems turn out to be very accurate. (That's good, or else why use simulations at all?)
For non-empirical models of chemical reaction kinetics, look up 'transition path sampling'. It's a way to sample the possible reaction mechanisms of a reaction and calculate kinetics. You need to perform a lot of sampling so it's currently limited to quickly occurring reactions (but it has recently worked on calculating the kinetics of small protein folding). The method does require that the reaction mechanism is known (ie. what the reaction coordinates are).
For the hybrid QM/MM stuff, look up 'Darrin York'. That stuff still needs some sort of reaction coordinate defined, but the simulations don't assume anything about the catalysis process itself. When you're working with thousands of atoms, it's hard to determine at which point did the substrate turn into product/etc. It's sometimes useful to create the reaction coordinate and use it as a 'slice' of the real data.
The methods are the same: splicing, mutating, and ligating DNA.
But the thought behind WHAT to splice, mutate, and ligate is very different. They use engineering and mathematics to _predict_ what will work _before_ they build it.
So it's not the trial and error prone protocol of traditional molecular biology. It's more like real engineering.
(Or, what, did you think they create a million different skyscrapers and the one which stays up is copied?);)
We do not have a model which reliably predicts some of the most simple chemical reactions, much less those in biochemistry.
That may have been true 30 years ago, but it's not now. In fact, we can predict the dynamics of biological processes, such as gene expression, signal transduction, and metabolism. The hard part about predicting these systems is that _there are so many components_.
Today, mechanical engineers can completely predict how a car with thousands of individual components will behave. Chemical engineers can completely predict how a huge factory with thousands of unit processes, each with their own highly complex physical behavior, will function.
In the future, biological engineers will be able to predict how a single cell with millions of components will behave over time.
Right now, we can only predict how small systems will work. But, as the methods get better over time and as enough information about those components is collected, the goal of predicting whole-cell behavior is clearly obtainable.
Well, the difference is the 'engineering' part. They're not cutting, splicing, or mutating large stretches of genes in the hope that it will confer some phenotype. They're using mathematics to _predict_ what will happen if they construct a specific stretch of DNA.
(I actually do research on the math part of this. And some on the building.)
The RTGs (which I mentioned) do not undergo the same reaction that nuclear fission power plants use to generate energy. It's not uranium or plutonium and the output is miniscule in comparison. That's what I meant.
Unsurprisingly enough, world didn't end, nor did it turn Canada into uninhabitable wasteland
If people get 10% more radiation than normal, get cancer, and sue the government for $2 billion, then the world won't end, but we'll have sooo much fun dealing with the consequences.
Well, true, but there's a lot more than temperature resistance that could be engineered into plants.
The biggest benefit would be to reduce the dependency on certain nutrients that may be scarce on Mars. Things like nitrates, sulphates, etc that would be difficult to bring along in sufficient quantities.
I think nuclear power is a clear necessity, but it has (so far) been a political brick wall. Absolutely no space ship has so far used nuclear fission. The fear is that the space ship will explode and rain down radioactive isotopes. (Of course, there's a lot of engineering that can prevent any catastrophic situations, but people don't care about that.) The closest thing is nuclear heating, which uses small quantities of radioactive isotopes to generate ~1kW of energy. Any Mars trip needs a few MW to do all of the things that need to be done.
Well, actually, if they grow the plant in a greenhouse on Mars then the ambient temperature will be pretty high. The sunlight on Mars is not filtered or blocked by a strong atmosphere and, just like greenhouses here on Earth, the energy will be mostly trapped inside. They can control the temperature by exposing un-insulated portions of the greenhouse to the cold outside (~-40 deg C in sunlight).
Slashdot Mirror
*Exactly*. And Intel's C/Fortran compiler is F A S T. (I use it on Itaniums as well. Not that Itaniums are all that fast, but who cares if you have 100+ of them. ;) )
For Linux, the Fortran compiler is also free right now (as in beer, for non-commercial use), which allows me to code and test from home on my x86 and then recompile it on Itanium. Easy!
I'm sorry, but you're wrong.
... a new JVM! So when the Cell processors are available for scientific computing, C and Fortran compilers will be written to take advantage of their amazing capabilities. How quickly do you think a JVM will be available??
Java performance depends entirely on the JVM. If the JVM sucks, then the Java program runs slow. The JVM may be good on PIV x86, but what about Power series? What about Itanium? Clusters don't often use x86's because they consume lots of power and require lots of cooling. And new computing architectures require
I read the website you cited and lot of the reasons for its 'Java is faster than C!' aren't true. Most of the reasons why Run-time compilation are complete BS. A good Fortran/C compiler will do the _exact same thing_. The website doesn't convince me. It has too few benchmarks and does not provide the code it uses. Not impressed.
So, for scientific computing, Java is still not the answer.
And, I'm not alone. If Java was faster, then don't you think IBM/Intel/SGI/HP would create nice JVMs on all of their HPC systems with nice development environments. They don't.
I completely agree, but you'd be surprised how many scientists (computer scientists even!) will complain that XYZ algorithm is too computationally intensive...and then you find out they implemented it in Java.
It's like "DUH!" C or Fortran. Pick one. Or C++, even. But Not Java.
Of course, when you broach the topic, you will hear things like 'C is old. Fortran is older!' Wtf cares? It's F A S T.
$48 million isn't much for multiple supercomputers.
;)
Did you look at the price tag for IBM's Blue Gene or Japan's Earth Simulator? Yeah...much more than $48 mil.
And I can't use either of those nice government funded ones! Bastards. They won't even accept an allocation request.
"But I only want to use 15 000 processors for an hour!" Heh.
It's an MPI environment. MPI stands for Message Passing Interface. It's a library of subroutines that allow you to send data from one processor to another. So if you wanted to process a 2 gig text processing, you could divide the text file into 500 chunks and have each processor perform some function on its chunk. Then you would collect all of the results together and save them.
Certain programs are more 'parallelizable' than others. The programs I run (and code) are very 'embarrassingly parallel'. That means that there is very little communication between each processor while the program is running. Certain programs require a large amount of communication between processors and a lot of special attention must be given to optimizing the architecture of the parallel program.
The bandwidth and latency between processors is pretty good, but no where near the actual speed of computing. So you need to balance the time it takes to send data from one processor to another and the time it takes to compute some results. Since the processor can do both at the same time (send/receive data while computing), it becomes a sort of art to synchronize it all together so that no processor is left idling with nothing to do.
Hope that explains it!
-Howard
it makes me smile.
... cool.
... the people of Slashdot!)
It's just so
(And the only people who I say that to are my research group members and
The TeraGrid is well managed too.. very few problems for such a huge system.
I was glossing over the differences (for the Slashdot crowd) between microRNA's, snRNAs, RNA interference, and other small RNA transcripts that bind to mRNA and perform X function.
I hadn't heard about RNA's regulating chromatin compaction. In yeast, is the mechanism known? Does it help catalyze acetylation? Does it inhibit histone binding to DNA?
It'll more likely be translational control via RNA.
RNA can quickly hybridize with regulatory regions of mRNA and change their translation rate.
And these RNA transcripts can be very small, but still regulate the translation of many genes. It'll be a while until the function of all of these RNA's are understood.
NetCDF is used more by non-computer scientists who need to store lots of data. It has an easier-to-use API. But, HDF5 can do more useful things. When NetCDF uses HDF5 as its underlying format, it'll get the best of both worlds: good API, good data structure.
When I started a software project about two years ago, I looked at both NetCDF and HDF5 for data formats. I chose NetCDF and have had zero problems (and it's been very easy to the software working nicely). I think using HDF would have added another 6 months of development time.
In the end, it won't matter: they'll both be equivalent.
I agree with a lot of your comments. There is a reason why online peer review is not going to happen for a long time.
;) Usually, they spend the time and come back with intelligent comments. Sometimes, not...but that can always happen either way.
But, for some fields, the first steps towards online peer review might be to open up commenting to online readers. Maybe the journal requires a sign in process. Maybe the journal allows anonymous commenting. Who knows. Clearly, there are problems to work out.
Also, if the field of research is small then the only problem is to make sure each interested professor is connected to the online peer review process. Social networking really isn't the problem: all you need is a invite/nominate system that encourages professors in similar fields to review each others' papers.
But, I think, a 'field' of research is not a concrete boundary. Scientific papers should be understandable by any professor in a similar field. I've often reviewed papers or grants that have very little to with my own research. That's fine...you just have to read into the subject a bit more. Likewise, my papers have been reviewed by people who have no clue what I am doing.
Well, I think the process will eventually evolve to take more advantage of the internet. For example, the PLOS journals and others are all freely available under the Creative Commons License. The review processes for many journals all take place over the internet and not through mail.
;)
How long will it take before a journal turns on online 'comments' below each article? Maybe you have to log in to reply, but it's a start. The next step would be to allow randomly selected members of an online journal to read a paper and review it. The step after that might be to designate a portion of the journal as "pre-published" and allow anyone to review it with only well-reviewed papers allowed to transfer to the "published" portion.
Of course, I'm just guessing. Credibility is the only asset a journal has. If they modify their review process that credibility may be damaged. Who knows..it might take new upstart journals to revolutionize the review process.
Though, imagine the Slashdot community trying to review a technical paper. Do you think that's a good idea??
The paper stating that ~50% of scientific papers are false is published in the Public Library of Science (PLOS) Medicine. The paper only examined medical studies and not scientific papers on physics, chemistry, engineering science, (and mathematics).
While molecular biology papers can be prone to statistically insignicant, but factually stated conclusions, the biggest culprits are clinical studies and 'large-scale' analyses of data.
Good experiments are constructed to give a 'yes' or 'no' answer based on the presence or absence of evidence. The zeal of high-throughput studies and analysis have put more pressure on good statistical analysis. Unfortunately, statistical analysis requires math...which sometimes eludes doctors and biologists. Hence, the problem of missuing statistics and stating inadequately supported conclusions.
-Howard
I hate people who say "It is a pity that ... " when they are flat out wrong. If you're going to be so condescending to someone, you better make sure you're absolutely correct. Otherwise, not only are you an idiot, but you're an arrogant idiot.
... " anyways?
Anyways, wtf says "It is a pity that
Do you even know what the word 'pity' means??
But looking in his JACS paper I see what you mean, he's doing a dynamics simulation.
Ah, gotcha. That does make more sense.
Right. But the inaccuracy of DFT methods is about 3-7 kcal/mol
I'm not sure what method was used, but I've seen errors of less than a kcal. You need the error to be that low to simulate biochemical reactions (which often have activation energies of only a few kcal). It might have been York's stuff, I forget. ('course, this isn't really my field.)
-Howard
It requires a full potential energy surface to work with. And that would be either empirical (MM, which is incapable of breaking/forming bonds) or non-empirical (QM, which is way too expensive to do any kind of PES scanning).
;)
I mentioned Darrin York because he recently gave a presentation on semi-empirical QM potentials. Basically, he performs hundreds of ab initio simulations of a certain type of reaction (he chose phosphorylations by a kinase protein) and then fits a semi-empirical model of the force field of the reaction. The potential is very accurate, but only for phosphorylation reactions. Then he performs QM/MM simulations using the semi-empirical force field, which is much faster now that the original ab initio calculations have already been completed.
That's wrong though. The transition-state is easily defined
That assumes there's only one transition state. Really, there's an ensemble of transition states where the potential energy is at a local maxima. The potential field in the large 6N-dim phase space is never very smooth, anyways, so there's always a ton of saddle nodes around the transition state 'region'.
Then again, there could be more than one 'region' of the phase space with local maxima. These are simply alternative mechanisms of the same reaction.
But I think you already know this and were simplifying it for me.
My advisor does MD simulations of protein-protein and protein-DNA interactions as well as kinetic Monte Carlo (mesoscale stochastic simulation) of large networks of biological components. (The latter is what I do.)
-Howard
I vote for blast too. :)
.... stochastic!".
;)
Although, even among my research colleagues, the running joke about my research is "It's
They're just jealous.
They're not revolutionizing the methods, per say. They're using them in different ways. For example, a group at MIT created 'BioBricks'. They are segments of pre-synthesized DNA sequences which can be ligated in any order you wish using the same exact procedure. The bricks are specially constructed (with specific restriction sites) so that the same protocol can be used over and over again to continually ligate together more and more BioBricks.
That way, you identify what you want to build and then you ligate together the bricks that contain each individual 'part'. They're 'replaceable parts'.
But they use the same ol' genetic engineering techniques to splice, mutate, and ligate those bricks: restriction sites, PCR, electrophoresis, antibiotic selection.
The methods that you listed are usually used for mammalian cells. Most of these guys use yeast or bacteria for their work. If you're working on bacteria, you can create single or multiple copy plasmids or integrate a plasmid into the bacteria's chromosome. If you're working on yeast, you can use homologous recombination to target the location of integration.
For bacterial and yeast cells, it's very well known how a gene is transcribed and translated into protein and what other proteins are involved. Because we know how the system generally works, we can accurately describe it using mathematics. Then we can use the mathematics to predict what will happen if we cut out portions of the system and paste them in different combinations.
The prediction part is the real revolution.
What I was talking about is the prediction of the what and not the how. For instance, given a DNA sequence for an enzyme, you cannot predict its structure. (yet)
... atoms. The current solution to that problem is to use a hybrid Quantum Mechanic / Molecular Mechanic (QM/MM) method that only uses ab initio calculations for the atoms located near the catalytic site and use regular molecular dynamic simulations for all other atoms. You need to merge the two together self-consistently (not easy), but people are working on it and making good progress.
;) )
/etc. It's sometimes useful to create the reaction coordinate and use it as a 'slice' of the real data.
Ok, yes, the protein folding problem (for any protein) is not solved yet. However, using transition path sampling, it's possible to fold a small 40 amino acid protein from a random structure. Google Folding@Home for details.
And even given the structure, you cannot predict the function an enzyme. (ab initio, anyway. Homology comparisons can help do it of course. But since we're talking 'deep understanding' here, I take it to mean ab initio)
Ab initio calculations are limited to ~200 atoms. Yes,
So, you're right when you say "you cannot make an ab-initio model", but if the end goal is to _predict_ the dynamics of a whole cell then one doesn't need to perform quantum calculations _on everything_. And the calculations don't need to be performed at the same time.
For example, you can take a single DNA sequence and run a simulation to predict its native structure. You can use homology (on both the sequence and structure) to estimate where the catalytic site is and what type of substrate likes to bind. Then you perform another simulation to predict the reaction mechanisms, kinetics, etc of that enzyme.
Then you repeat that process for every protein in a single cell. (Leave, grab a cup of coffee, spawn four generation of children, come back, get the results.
Then you take all of that data and use a different type of simulation method to predict what will happen when all of those proteins are thrown together. You don't need to use ab initio to do that (or molecular dynamics simulations).
Now, of course, there's experimental data out there so we don't need to in silico predict _everything_. We can jump start the process of predicting the dynamics of biological systems by only using well-characterized proteins and genes. And the mesoscale-type simulations of those systems turn out to be very accurate. (That's good, or else why use simulations at all?)
For non-empirical models of chemical reaction kinetics, look up 'transition path sampling'. It's a way to sample the possible reaction mechanisms of a reaction and calculate kinetics. You need to perform a lot of sampling so it's currently limited to quickly occurring reactions (but it has recently worked on calculating the kinetics of small protein folding). The method does require that the reaction mechanism is known (ie. what the reaction coordinates are).
For the hybrid QM/MM stuff, look up 'Darrin York'.
That stuff still needs some sort of reaction coordinate defined, but the simulations don't assume anything about the catalysis process itself. When you're working with thousands of atoms, it's hard to determine at which point did the substrate turn into product
-Howard
The methods are the same: splicing, mutating, and ligating DNA.
;)
But the thought behind WHAT to splice, mutate, and ligate is very different. They use engineering and mathematics to _predict_ what will work _before_ they build it.
So it's not the trial and error prone protocol of traditional molecular biology. It's more like real engineering.
(Or, what, did you think they create a million different skyscrapers and the one which stays up is copied?)
We do not have a model which reliably predicts some of the most simple chemical reactions, much less those in biochemistry.
That may have been true 30 years ago, but it's not now. In fact, we can predict the dynamics of biological processes, such as gene expression, signal transduction, and metabolism. The hard part about predicting these systems is that _there are so many components_.
Today, mechanical engineers can completely predict how a car with thousands of individual components will behave. Chemical engineers can completely predict how a huge factory with thousands of unit processes, each with their own highly complex physical behavior, will function.
In the future, biological engineers will be able to predict how a single cell with millions of components will behave over time.
Right now, we can only predict how small systems will work. But, as the methods get better over time and as enough information about those components is collected, the goal of predicting whole-cell behavior is clearly obtainable.
(This is my research.)
-Howard
Well, the difference is the 'engineering' part. They're not cutting, splicing, or mutating large stretches of genes in the hope that it will confer some phenotype. They're using mathematics to _predict_ what will happen if they construct a specific stretch of DNA.
(I actually do research on the math part of this. And some on the building.)
-Howard
Uh, what about the Itaniums?
They're 64-bit and Intel has sold quite a few.
The RTGs (which I mentioned) do not undergo the same reaction that nuclear fission power plants use to generate energy. It's not uranium or plutonium and the output is miniscule in comparison. That's what I meant.
Unsurprisingly enough, world didn't end, nor did it turn Canada into uninhabitable wasteland
If people get 10% more radiation than normal, get cancer, and sue the government for $2 billion, then the world won't end, but we'll have sooo much fun dealing with the consequences.
Well, true, but there's a lot more than temperature resistance that could be engineered into plants.
The biggest benefit would be to reduce the dependency on certain nutrients that may be scarce on Mars. Things like nitrates, sulphates, etc that would be difficult to bring along in sufficient quantities.
I think nuclear power is a clear necessity, but it has (so far) been a political brick wall. Absolutely no space ship has so far used nuclear fission. The fear is that the space ship will explode and rain down radioactive isotopes. (Of course, there's a lot of engineering that can prevent any catastrophic situations, but people don't care about that.) The closest thing is nuclear heating, which uses small quantities of radioactive isotopes to generate ~1kW of energy. Any Mars trip needs a few MW to do all of the things that need to be done.
Well, actually, if they grow the plant in a greenhouse on Mars then the ambient temperature will be pretty high. The sunlight on Mars is not filtered or blocked by a strong atmosphere and, just like greenhouses here on Earth, the energy will be mostly trapped inside. They can control the temperature by exposing un-insulated portions of the greenhouse to the cold outside (~-40 deg C in sunlight).
The biggest difficulty is getting enough water.
-Howard