Computer simulations. How do they support theory?

[Enchilada]

Originally Posted by DJDD

good points raised earlier.
would you expand upon your comments that cloning and stem cell (research, I presume) is "non-scientific biology"?

No I would expand, considering the edicts of the forum here. Clearly I am speaking of those against such sciences, believing that such research should not be done at all based on "non-scientific grounds" I.e. Religious dogma

I actually meant against cloning and stem cell. I corrected the previous post as well!!

[SMR (Steve)]

Hi Carl

Originally Posted by renormalised

Be very careful about using computers ...like all tools, they're only as good as those that use them.

In my experience, most tools are much better than the fools using them ...

To me, Rob seemed to be saying that these cosmology guys playing with computers in Granada were fools who had no idea about complex ideas like GIGO. Do you agree?

Originally Posted by renormalised

What they're great for is [...] constructing representations of theoretical models

Which is exactly the thing that Rob was complaining about -- using computers to represent/simulate theoretical models is dangerous. You're saying they're "great". So how come you guys are agreeing?? You should be at each other's throats!!

BTW, my guess is that you haven't written any code either, right?

Steve.

[renormalised (Carl)]

Originally Posted by SMR

Hi Carl



In my experience, most tools are much better than the fools using them ...

To me, Rob seemed to be saying that these cosmology guys playing with computers in Granada were fools who had no idea about complex ideas like GIGO. Do you agree?



Which is exactly the thing that Rob was complaining about -- using computers to represent/simulate theoretical models is dangerous. You're saying they're "great". So how come you guys are agreeing?? You should be at each other's throats!!

BTW, my guess is that you haven't written any code either, right?

Steve.

True...a tool can't do anything until it's used. if the person using the tool isn't up to scratch, so to speak, then the tool being used is either misused or not used to its full potential.

Nothing complex about GIGO...you get out of something only that which you put in. If what you put in is incomplete, then what you get out is likewise. If it's erroneous, then you generate errors. What Rob was saying is that they don't know enough about the subject they were modeling (dark matter) to get out of the simulations anything which would be really meaningful, since their initial input was incomplete in nature and possibly erroneous to begin with. All they would get would be a model that would only confirm what they already knew, or thought they knew. A model, by the way, which rests on not very much empirical evidence at all, other than an observed gravitational effect they don't fully understand in the first place.

It can be "dangerous" if they're applied to critical situations where life and limb are at stake and the models turn out to be wrong to begin with. How often has that happened in industry and engineering situations. Quite often, unfortunately.

As for not writing any code...sorry to disappoint you there. I've written code in basic, fortran, pascal and C, plus used SQL as well. I know how complex it can get and that mistakes are all too common, no matter how careful you try to be. But that doesn't make any theory you may use the simulations/programs to visualise any more correct or complete than beforehand, or after. The simulations don't do anything more than act as a way to see what you theorised about. They're great for creating visualisations of what you're trying to express in theory, but that's about it. For example, an N-body simulation will build you a solar system no problems, but it doesn't build you a replica of what you're looking at. That requires variables most of which you cannot predict from 1st principles. They have to come via observation and experience...each solar system has unique properties which are due to factors specific to their formation, not repeatable elsewhere. Models, by their nature, are generalisations which should be used only as guides to further study, not as a means to an end in themselves.

[Enchilada]

Originally Posted by renormalised

As for not writing any code...sorry to disappoint you there. I've written code in basic, fortran, pascal and C, plus used SQL as well. I know how complex it can get and that mistakes are all too common, no matter how careful you try to be. But that doesn't make any theory you may use the simulations/programs to visualise any more correct or complete than beforehand, or after. The simulations don't do anything more than act as a way to see what you theorised about. They're great for creating visualisations of what you're trying to express in theory, but that's about it. For example, an N-body simulation will build you a solar system no problems, but it doesn't build you a replica of what you're looking at. That requires variables most of which you cannot predict from 1st principles. They have to come via observation and experience...each solar system has unique properties which are due to factors specific to their formation, not repeatable elsewhere. Models, by their nature, are generalisations which should be used only as guides to further study, not as a means to an end in themselves.

Hence posting the link to simulation/ models in the on line programs "N-body 4 and 6", which clearly demonstrates what your saying!!

[Kal (Andrew)]

For many years I participated in a distributed computing project called Folding@home (f@h) which simulated the folding of proteins to help find cures for diseases such as Alzheimers.

Here is one of the published articles they wrote titled "HOW WELL CAN SIMULATION PREDICT PROTEIN FOLDING KINETICS AND THERMODYNAMICS?"

Too long for me to read, but their conclusions are interesting:

CONCLUSIONS
In the end, an understanding of complex biophysical phenomena will require computer simulation at some level. Most likely, experimental methods will never yield the level of detail that can today be reached with computer simulations. However, the great challenge for simulations is to prove their validity. Thus, it is naturally the combination of powerful simulations with quantitative experimental validation that will elucidate the nature of how proteins fold. How close are we to achieving this goal? In many ways, there has been great progress. The ability to quantitatively predict rates, free energies, and structure from simulations on the basis of physical force fields reflects significant progress made over the past five years. It also draws attention to a new challenge. Even the prediction of experimental observables, such as rates, within experimental uncertainty does not prove that the simulations will yield correct insights into the mechanism of folding. Indeed, recent work suggests that computational models can both agree with experiment and disagree with each other (89). We must therefore push the link between simulation and experiment further by connecting the two with new observables, multiple techniques, and increasingly strict quantitative comparison and validation of simulation methods.Without more detailed experiments, we may not be able to sufficiently test current simulation methodology and the trustworthiness of refined simulations may remain unclear. Nonetheless, the ability to predict rates, free energies, and structure of small proteins is a significant advance for simulation, likely heralding even more significant advances over the next five years.

[Robh (Rob)]

Originally Posted by SMR

Hi Rob

It's pretty clear that you don't know what a computer program is. Written much code recently?

... Or are you just really looking for a way to discredit ideas that you don't understand, and extend that to other topics that you feel political about? Honest answer, please.

Thanks!

Steve.

Hi Steve,

In an attempt to clarify my stance, I am going to go over some past territory again. In my thread starter, I stated "How can a simulation support a particular theory? I suspect that the verification of a theory can only occur if a simulation modeling it makes an observable prediction which can be confirmed. My worry is that many simulations are scaffolding an imaginary Universe without any real-world confirmation." This is my personal observation and my opinion. It was an expression of concern. The article I chose was for a particular reason. I thought it was interesting because a model had been built on how dark matter is distributed in galaxies based on assumptions about dark matter, of which we know little and which may or may not exist. I assumed some computer simulation was involved here but I may be wrong about this and have stated so. Irrespective, the question about how simulations are validated is open to discussion.

Note, that I did not say that all simulations are questionable. Neither was I attempting to ridicule or discredit any of the research. It was a question of concern. If the model produced doesn't confirm or make a prediction about the real world, how is it validated?

In answer to one of your questions, I have experience programming at the machine level and also in languages such as FORTRAN, Pascal and BASIC. I have used and am currently using code to produce star maps on my own computer. I have done some computer simulations myself as a teacher but I don't claim to be an expert in the area.

In my thread starter, I finished with "Anyone else with any thoughts on this?". I was actually welcoming other people's views on this. And I welcome yours. I don't think I was being dogmatic in any way.

Regards, Rob.

[Robh (Rob)]

Originally Posted by Kal

For many years I participated in a distributed computing project called Folding@home (f@h) which simulated the folding of proteins to help find cures for diseases such as Alzheimers.

Here is one of the published articles they wrote titled "HOW WELL CAN SIMULATION PREDICT PROTEIN FOLDING KINETICS AND THERMODYNAMICS?"

Too long for me to read, but their conclusions are interesting:

CONCLUSIONS
In the end, an understanding of complex biophysical phenomena will require computer simulation at some level. Most likely, experimental methods will never yield the level of detail that can today be reached with computer simulations. However, the great challenge for simulations is to prove their validity. Thus, it is naturally the combination of powerful simulations with quantitative experimental validation that will elucidate the nature of how proteins fold. How close are we to achieving this goal? In many ways, there has been great progress. The ability to quantitatively predict rates, free energies, and structure from simulations on the basis of physical force fields reflects significant progress made over the past five years. It also draws attention to a new challenge. Even the prediction of experimental observables, such as rates, within experimental uncertainty does not prove that the simulations will yield correct insights into the mechanism of folding. Indeed, recent work suggests that computational models can both agree with experiment and disagree with each other (89). We must therefore push the link between simulation and experiment further by connecting the two with new observables, multiple techniques, and increasingly strict quantitative comparison and validation of simulation methods.Without more detailed experiments, we may not be able to sufficiently test current simulation methodology and the trustworthiness of refined simulations may remain unclear. Nonetheless, the ability to predict rates, free energies, and structure of small proteins is a significant advance for simulation, likely heralding even more significant advances over the next five years.

Thanks Andrew,

I found this very interesting and enlightening!

Regards, Rob.

[SMR (Steve)]

Hi Carl

Originally Posted by renormalised

As for not writing any code...sorry to disappoint you there. I've written code in basic, fortran, pascal and C, plus used SQL as well.

Then I'm happy to admit that my guess was completely incorrect.

Steve