I will admit that even though I was a comp sci major at university, biochemistry has always been a bit of an interest.

There are a few questions that I'd like to try to get answers to, one of which has puzzled me for quite some time, while the others have arisen from reading about what Rosetta does.

Long term puzzle first. Take an amino acid like alanine. It has two stereo-isomers, because if you view the CH3, the H, the NH2 and the CO2H that are joined to the central C as the points of a tetrahedron, and you place the CH3 as the top point of the tetrahedron, looking down on it from above there are two ways you can arrange the H/NH2/CO2H around the bottom three points. OK so far?

However, there are a couple of others (isoleucine and threonine in particular) that appear to have a second C in them that has the same property: four different radicles bonded to it, thus allowing a second set of stereo-isomerism. So what about these two guys. I know that for the remaining 17 that have "one way stereo-isomerism" we (and every other life form on planet Earth) only ever use of the two isomers. What about these two? Are there a total of four isomers? Do we only use one? What's the story?

Rosetta questions now. In another post by David Baker, he described the folding of a protein, and the energy of each particular fold as being like trying to map a 500 dimensional space. Surprisingly, that's a concept that I can grasp quite readily, but it gives rise to some questions.

Am I correct in thinking that as the protien folds, it doesn't move smoothly along the dimensions in that space, but has a tendancy to "snap" to discreet points. If we consider each dimension as measuring the angle that a particular rotatable bond holds, that bond (as far as I understand) tends to rotate by discreet angles (i.e. snap). This means that potentially, if we move from one point in the graph to a neighbor directly along a single dimension, that means a single rotation of a bond. However, if that bond is between a C and a CO2H right in the middle if the protein chain, it could throw half the chain way off into the middle of nowhere, meaning it jumps from a fairly low energy state to a much higher energy state (or vice versa).

Speaking mathematically, does this mean that there are tremendous discontinuities in the shape of the map?

Also, going back to how amino acids join. Take something like phenylalanine. It was noted elsewhere that each amino acid has an average of five rotatable bonds. I'd take any odds you want to give that the benzene ring on the end of phenylalanine can rotate with respect to the rest of the acid, on an axis that lies along the bond to the CH2 that joins it to the rest.

Working with this guy for a moment, let's say we look at the following three bonds (which I'm assuming are rotatable): C-NH2, C-CO2H, the bond between my NH2 and the next guys CO2H. Twisting any of these falls into the class above, it could have very far reaching effects. However, if we twist the Benzene ring, I don't immediately see how it can have much effect, since it's a very localized change. Is this true?

Where I'm going is this. When you're doing the initial folding tests (described as the protein "flailing around wildly"), do you do any sort of prioritization? If not, have you thought about something akin to this. Dreaming up a metric out of thin air (tune it as necessary), try assigning a weight to each bond. You can sum up the total atomic weights on either side of the bond, my gut feel is to take the product of these two weights. Then spend more time adjusting the heavier weighted bonds first.

P.S. Don't assume I know THAT much about this stuff. I fall into that wonderful class of people who know just enough to be incredibly dangerous. :)

Edit - watch what you're typing, it's phenylalanine, not phylalanine.

---

I'm not part of the Rosetta team, but I'm a biochem student, so I can answer the straight biochem stuff....

Long term puzzle first. Take an amino acid like alanine. It has two stereo-isomers, because if you view the CH3, the H, the NH2 and the CO2H that are joined to the central C as the points of a tetrahedron, and you place the CH3 as the top point of the tetrahedron, looking down on it from above there are two ways you can arrange the H/NH2/CO2H around the bottom three points. OK so far?

However, there are a couple of others (isoleucine and threonine in particular) that appear to have a second C in them that has the same property: four different radicals bonded to it, thus allowing a second set of stereo-isomerism. So what about these two guys. I know that for the remaining 17 that have "one way stereo-isomerism" we (and every other life form on planet Earth) only ever use of the two isomers. What about these two? Are there a total of four isomers? Do we only use one? What's the story?

I have a professor at my university whose favourite line is "life is stereospecific". In other words, although there are 2 isoforms around the alpha carbon (that's the one in the middle), 99% of all organisms only use one form of amino acid, the "L" form, similarly, we only use one specific chirality of threonine and isoleucine. (There are some bacteria that use the D amino acids as a form of protection in their outer walls, because no other organisms have the enzymes to cut up those proteins!!)

Also, going back to how amino acids join...
[snip]
Working with this guy for a moment, let's say we look at the following three bonds (which I'm assuming are rotatable): C-NH2, C-CO2H, the bond between my NH2 and the next guys CO2H. Twisting any of these falls into the class above, it could have very far reaching effects. However, if we twist the Benzene ring, I don't immediately see how it can have much effect, since it's a very localized change. Is this true?

You're not quite right on this one. The NH-COO bond (which is between the N and one of the oxygens)is not rotatable because it is what is called a "partial double bond", which basically does not rotate.

[img=http://dis.shef.ac.uk/ruth/pep.gif]

In terms of if rotating the benzene ring does much... I'd say not so much, because a benzene ring is fairly rigid. However, if you're rotating a glutamic acid or leucine, which are amino acids with long side chains, it could make a big difference, especially for the acidic and basic side chains, which are often involved in the functionality of the protein.

I hope that helps a little on the biochem of the matter...

---

Questions? Try the Wiki!
Take a look inside my brain

Long term puzzle first. Take an amino acid like alanine. It has two stereo-isomers, because if you view the CH3, the H, the NH2 and the CO2H that are joined to the central C as the points of a tetrahedron, and you place the CH3 as the top point of the tetrahedron, looking down on it from above there are two ways you can arrange the H/NH2/CO2H around the bottom three points. OK so far?

However, there are a couple of others (isoleucine and threonine in particular) that appear to have a second C in them that has the same property: four different radicles bonded to it, thus allowing a second set of stereo-isomerism. So what about these two guys. I know that for the remaining 17 that have "one way stereo-isomerism" we (and every other life form on planet Earth) only ever use of the two isomers. What about these two? Are there a total of four isomers? Do we only use one? What's the story?

Only one isomer of each amino acid is used biologically, including the chiral centers of the side chains.

Rosetta questions now. In another post by David Baker, he described the folding of a protein, and the energy of each particular fold as being like trying to map a 500 dimensional space. Surprisingly, that's a concept that I can grasp quite readily, but it gives rise to some questions.

Am I correct in thinking that as the protien folds, it doesn't move smoothly along the dimensions in that space, but has a tendancy to "snap" to discreet points. If we consider each dimension as measuring the angle that a particular rotatable bond holds, that bond (as far as I understand) tends to rotate by discreet angles (i.e. snap). This means that potentially, if we move from one point in the graph to a neighbor directly along a single dimension, that means a single rotation of a bond. However, if that bond is between a C and a CO2H right in the middle if the protein chain, it could throw half the chain way off into the middle of nowhere, meaning it jumps from a fairly low energy state to a much higher energy state (or vice versa).

Speaking mathematically, does this mean that there are tremendous discontinuities in the shape of the map?

This is true, and a fair amount of our research goes into figuring out ways to move around sensibly in this high dimensional space. Besides the lever arm sensitivity that you describe here, there are also discontinuities due to the fact that atoms cannot be in the same place at the same time. Small moves can result in very large changes in energy.

Also, going back to how amino acids join. Take something like phenylalanine. It was noted elsewhere that each amino acid has an average of five rotatable bonds. I'd take any odds you want to give that the benzene ring on the end of phenylalanine can rotate with respect to the rest of the acid, on an axis that lies along the bond to the CH2 that joins it to the rest.

Working with this guy for a moment, let's say we look at the following three bonds (which I'm assuming are rotatable): C-NH2, C-CO2H, the bond between my NH2 and the next guys CO2H. Twisting any of these falls into the class above, it could have very far reaching effects. However, if we twist the Benzene ring, I don't immediately see how it can have much effect, since it's a very localized change. Is this true?

Where I'm going is this. When you're doing the initial folding tests (described as the protein "flailing around wildly"), do you do any sort of prioritization? If not, have you thought about something akin to this. Dreaming up a metric out of thin air (tune it as necessary), try assigning a weight to each bond. You can sum up the total atomic weights on either side of the bond, my gut feel is to take the product of these two weights. Then spend more time adjusting the heavier weighted bonds first.

In fact, our algorithm works much like this. You can think of a protein as made up of a backbone and side chains. The backbone contains those rotatable bonds can have far reaching effects in the overall shape of the chain. The side chains are the parts of the chain with rotable bonds that have relatively little effect on the overall shape. In the inititial "flailing" stage, only backbone degrees of freedom are sampled. The side chains are represented as a sort of averaged out blob. It is in the second stage of the algorithm that the side chain degrees of freedom are sampled.

P.S. Don't assume I know THAT much about this stuff. I fall into that wonderful class of people who know just enough to be incredibly dangerous. :)

It seems that the least we can do is answer your questions about the program that you are generously running on your computers.

---

I didn't ask the question, but thank you for the answer, Jack Schonbrun.

I'm going to love this place.

---

I didn't ask the question, but thank you for the answer, Jack Schonbrun.

I'm going to love this place.

I'm happy to provide info when I can. Recommend us to your friends!

Jack

---

I didn't ask the question, but thank you for the answer, Jack Schonbrun.

I'm going to love this place.

I'm happy to provide info when I can. Recommend us to your friends!

Jack

Thanks to both you and eberndl for clearing things up. And yeah, I'm going to like this place too. :)

---