23 replies to this topic

[Mind]

Vijay Pande!

Folding@Home.

More details to come.

[Mind]

In case you are not already aware, the Pande Lab is the founding scientific group of Folding@Home - a distributed computing project dedicated to solve the mysteries of protein folding and conquer human diseases. Folding@Home is the largest supercomputer in the world, soon to cross the 5 petaflop barrier. Folding@Home Homepage.

The Immortality Institute is heavily involved in promoting Folding@Home through the F@H prize. Check our folding forum for all the latest.

This week on the Sunday Evening Update, the founder himself - Vijay Pande - will make an appearance. It is your opportunity to ask questions and find out the latest news and research.

Attached Files

•  pandeteach2.jpg   11.7KB   11 downloads

Edited by Mind, 14 January 2009 - 07:56 PM.

[brokenportal]

This should be excellent. We should try to make sure to get the question right about asking him if we can change the teams name for the cause. Maybe we should try to get him to support the cause quick, not saying it would work, but try and then express our concern about adding .com to longevitymeme and ask him if there are any circumstances it could be done under. Like, if we get a certain amount of points or something would he consider it etc...

[Mind]

Brokenportal is talking about this discussion. I don't think it is a big problem and that the current name is sufficient. Also, Vijay is more of the science leader of Folding@Home. I don't think he has much to do with the computer/technical/competition stuff.

[Mind]

Here is a hint of things to come from folding@home: folding finds small molecules that could help cure Alzheimers.

This is something I hope to delve into during the interview. What other research is going on with regards to diseases and aging? Also, I would like to have Pande speculate on what type of research is possible as folding@home passes 10, 20, 100 petaflops.

I would very much appreciate some questions from anyone who has more in depth knowledge about protein folding, physics, and biochemistry. If you can't attend the show live here on Sunday night at 6pm EST, then please ask your detailed question here in the forum.

[Mind]

One question I would like to ask is what the greatest challenges will be in the next couple of years. Are they having any hardware or bandwidth issues as more petaflops come online?

[Heliotrope]

When will this project finish? Human Genome Project finished a couple yrs early (well within a decade), under budget, etc How many years for FAH?

There's a limited # of proteins. What happens after everything is computed?

[dnamechanic]

... Are they having any hardware or bandwidth issues as more petaflops come online?

MichaelGR posted a link leading to an excellent video.

The video shows Professor Pande making an informative presentation that answers many questions about F@h.

A brief summary of some the topics covered can be found here.

[brokenportal]

I want to know if he thinks that we can work with biology to stop, get around or slow down aging and what he sees as the causes biggest road block to public acceptance.

Im curious about HYPs question too. I never really thought about them running out of proteins to fold. Will they, and do they have other projects lined up to take its place? Maybe there would be room down the road to negotiate for the unlimited lifespans cause here to take over that computing power for projects.

[niner]

I want to know if he thinks that we can work with biology to stop, get around or slow down aging and what he sees as the causes biggest road block to public acceptance.

I really don't think he's going to have any thoughts on this; it's just entirely outside of his field. Maybe by being here and talking to us, he'll get introduced to the concept, and that would be good.

Im curious about HYPs question too. I never really thought about them running out of proteins to fold. Will they, and do they have other projects lined up to take its place? Maybe there would be room down the road to negotiate for the unlimited lifespans cause here to take over that computing power for projects.

This isn't going to happen in the forseeable future. They could fold the same little protein thousands of times while tuning small details of the methodology. They could simulate the same protein with 100,000 or a million small molecule ligands in a virtual screening experiment. (i.e, a million different simulations) They could look at supramolecular assemblies of any number of proteins, or simulate a protein embedded in a lipid membrane. There's enough stuff to simulate that they could keep busy for decades.

[Heliotrope]

. There's enough stuff to simulate that they could keep busy for decades.

decades, or maybe a century of fah? if strong AI and super-super-computers help, then maybe soon

[Heliotrope]

impressed with Mind's ability to get top ppl on show. How'd you convince them? Do you just email + make cold calls and sign off as "exe dir." of II ?

many serious scientists would consider LE or "imm." as pipe-dream projects , not realistic, death defeats all . etc. Some even signed a collective letter saying "it should not have serious discussions in academic circles" why do we think we're diff? Every generation died, was there even a single person in that past that escaped grim reaper's fingers, the ultimate law in universe? No. They all died/die.

Edited by HYP86, 18 January 2009 - 12:25 AM.

[Heliotrope]

ask what Pande's position is on LE. Does he believe in this? ask how much effort or results of FAH is directed there.

What'll FAH look like years down the road? does he think AI'll help?

[niner]

Questions for Dr. Pande:

Where do they stand on being able to go from sequence to structure for an unknown protein? Have they entered any of the CASP competitions, or performed similar tests on their own?

If they are not yet able to reliably fold proteins, where is improvement needed? Force fields? Solvation? Electrostatics? Time Scales? Other?

What sort of solvation models are they using? Periodic Boundary Conditions? Soft Sphere? Blob?

Do they add ions to obtain charge balance in their simulations?

What electrostatic models do they think work the best for protein folding?

Can they reliably model the binding mode of a small molecule ligand?

[Heliotrope]

will they also try things like simulating gene interactions? ex. Do similar things with nucleic acids nitrogenous bases

will they try to use computer pwer for other molecules/compounds? do simulation of carbohydrate, fats, design /exp. drug/med interactions. will they try to create and build new things


what's Pande's future plan? for fah? for himself, other than teach, research, cure disease?

Edited by HYP86, 18 January 2009 - 12:52 AM.

[Mind]

Seems I forgot to link to the channel. Sunday Evening Update.

There is a text chat at the channel where you can converse with the rest of the audience and ask questions during the interview.

[brokenportal]

I want to know if he thinks that we can work with biology to stop, get around or slow down aging and what he sees as the causes biggest road block to public acceptance.

I really don't think he's going to have any thoughts on this; it's just entirely outside of his field. Maybe by being here and talking to us, he'll get introduced to the concept, and that would be good.

Im curious about HYPs question too. I never really thought about them running out of proteins to fold. Will they, and do they have other projects lined up to take its place? Maybe there would be room down the road to negotiate for the unlimited lifespans cause here to take over that computing power for projects.

This isn't going to happen in the forseeable future. They could fold the same little protein thousands of times while tuning small details of the methodology. They could simulate the same protein with 100,000 or a million small molecule ligands in a virtual screening experiment. (i.e, a million different simulations) They could look at supramolecular assemblies of any number of proteins, or simulate a protein embedded in a lipid membrane. There's enough stuff to simulate that they could keep busy for decades.

To both of these responses I would say, you never know. A part of our business is the meming business after all. I wouldnt grill him about it either if thats what your saying.

[Heliotrope]

won't make to the show, so Q: his predictions for FAH for far future. what'll else will it calculate besides folding protein

[AgeVivo]

I'd be VERY interested to know if he ever discussed with colleagues whether there could be potential programs for life extension??
- i mean more specificly anti-aging that protein folding.

I'm also interested in knowing what the program basically does on our computers for any given protein:
- start from the [sequence | 2ndary strcuture] in a non-folded position?
- then do simulations of various foldings [2d? | 3d?]?
- guided by [energy functions | comparison with known foldings of similar proteins] ?

I'm also interested in the algorithm, before even considering parallelisation of computers:
- algorithm: i understand that for a given protein the programs simulates many possible evolutions of folding, which is very slow because time must be VERY much discretized and because MANY simulations must be done. Aren't there faster alternatives, such as directly finding stable foldings by dynamic programming? Could he indicate us one or a few simple papers to understand the main algorithm/energy functions?
- are GPUs currently used by the program? what is in practice the computational-gain: 100? 1000?
- is adaptive sampling (more efficient simulation of possible foldings) currently used? gain: 100? 1000?
- other tricks?

Now on the parallel programming part (obviously his field):
- how scalable is it?
- why openMM rather that MPI?
- programming language for GPUs: CUDA currently?, what does he think of openCL?

Edited by AgeVivo, 18 January 2009 - 07:01 PM.

[Mind]

Starting in 30 minutes

[brokenportal]

Show is on now.

[Mind]

Direct link to the video here

It was a great interview. Dr. Pande was well prepared to answer a variety of questions. Technical aspects; audio and video were superb.

I wasn't able to get to all of the questions. A couple of them were a little over my head and I was afraid I would not be able to adequately explain them.

He said they are experimenting with some new software and that a new SMP client will be available before the end of the year.

He said folding@home comes pre-loaded on PS3s and ATI graphics cards, and that they are in communication with other vendors/platform makers to get it pre-installed but nothing concrete as of yet.

He explained that one work unit is typically a nano-second or two, but that a GPU does about 100 nanoseconds of a fold.

As far as getting out into the cloud for storage and analysis (hardly any local processing at Stanford) he said that move would come in a couple of years. He expects that they have enough storage space and bandwidth to last that long.

He is not sure if the current Alzheimer's results will be applicable to AGEs or other aspects of aging.

They have not competed in any CASP competitions but are interested.

Edited by Mind, 19 January 2009 - 01:49 AM.

[Mind]

Hey Niner, I did ask the question about binding modes but I can't quite remember his exact answer. Sorry, you'll have to listen. The question probably occurs about 15 to 20 minutes in.

[niner]

Hey Niner, I did ask the question about binding modes but I can't quite remember his exact answer. Sorry, you'll have to listen. The question probably occurs about 15 to 20 minutes in.

Yeah, I remember that. He said they had looked at not only binding mode but also affinity. To get the binding affinity, or "tightness" of binding, you need to determine the free energy of association of the ligand and receptor. It's a more sophisticated calculation than just calculating a trajectory. I'd like to learn more about how their results compare to experiment. That was the motivation for the CASP question; can they actually determine a correct structure? It's kind of the rubber meeting the road. I'll probably need to dig up some of their papers, but (oddly enough) the computational chemistry literature has been a little slow to go open access.