Enfuvirtide is a small (36 residue), human designed protein used for the treatment of HIV AIDS (effective against HIV-1 only). It is currently the only US-FDA approved drug in its class and also (rightly or wrongly) the most expensive at about US$25,000 per year.

It is a fusion inhibitor and works because HIV viruses have to dock at a particular site on a cell in order to enter the cell and reproduce; the shape of this drug/protein prevents the docking from occurring thus protecting the cell.

Now from what I understand, patents on naturally occuring proteins are not allowed. However, this protein is not natural--it is a "rationally designed" protein. Is it able to be patented?

Even if it is patented, shouldn't it be relatively easy to replace a few amino acids while retaining the shape and function of this drug/protein? Afterall, many proteins have share the same shape but have divergent sequences. Would doing so allow the creation of a different drug not subject to the existing patent?

And in general, how easy is it to dermine the effect of replacing a given residue with another on a protein of known shape? It seems like there would be some obvious "critical" residues (e.g. a Cysteine participating in a disulfide bridge or a Glycine in a tight turn where nothing else will fit). But it also seems there would be some easily-determined likely candidates for replacement (replacing one hydrophobic residue of similar size with another on the interior of a globular protein, perhaps).

Is this easy to dermine in general, or only for some special cases? If this is easy to determine, couldn't you analyze proteins of known shapes and come up with a sort of "regular expression" that describes it which would allow you to perform searches for similarly shaped proteins, even if their sequences aren't very similar.

For example, say you have protein-A and you know its shape. It's sequence is:
Ala-Gly-Leu-Thr-Val

You analyze the critical amino acids and come up with a regular expression like this:

[Any hydrophobic]-Gly-[Leu OR Lys OR Val]-[Any hydrophilic]-[Anything except Pro]

and store both in a database.

Later, you're interested in protein-B, which you do not know the shape of.

It's sequence is:
Met-Gly-Val-Asn-Ser

If you compare sequences only, they don't appear closely related (only 1 in 5 match), but if you compare it to the regular expression, it does suggest a similarity. Obviously, real proteins would be longer and the regular expressions might be much more restrictive, but hopefully that gives a feel for what I'm thinking. Does something like this already exist?

---

You may want to ask this over on the boards at SIMAP where their second application, (hmmer - not released yet), sounds like it may be of interest to you. "hmmer" records "functional attributes of sequences".

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

You may want to ask this over on the boards at SIMAP where their second application, (hmmer - not released yet), sounds like it may be of interest to you. "hmmer" records "functional attributes of sequences".

Thanks for the pointer. I didn't find a whole lot of information (at least, not in English and my high-school German doesn't help much) but SIMAP sounds interesting. However, I did read that the "hmm" in "hmmer" stands for Hidden Markov Model which implies that its a neural-net sort of approach...not really what I was thinking about below.

---

Hi Frank,

what you aim at is not easy to do. The problem ist that there are so many proteins with different kinds of tasks and properties. You can pretty easily replace single amino acids with others of similar properties. In many cases that does not damage the functionality too much.

Even the genetic code takes this possible replacement into account: similar amino acids often share "regions" of the code if you write it down systematically, meaning that if a mutation takes place its consequences are somewhat mitigated.

However, the effect of replacing one residue by another cannot be easily judged from the sequence itself, as you need much more information. E.g. where that residue will be in the folded protein: is it inside a fat-like part and just part of a "lump" that stabilizes the backbone's shape? Or is it rather part of an active site and may be responsible for cracking up or synthethising very small molcules? In the latter case the shape of the substrates may have to match so closely that the effect is devastating, in the first case you may get away with practically no change at all.

There are many proteins with absolutely different sequence and similar functionality and vice versa - sequence alone tells only part of the story. That is one of the reasons why Rosetta is done: You never can tell the function unless you know the shape.

Of course a lot of time is spent on systematics and it has its own right to get an overview over the jungle of different protein shapes, sequences and tasks. But it is all rather an encyclopedic approach and one for helping to establish phylogenetic trees. Making it an exact and predictable science is the only way to put it on solid ground.

Hi Christoph,

However, the effect of replacing one residue by another cannot be easily judged from the sequence itself, as you need much more information.

But what if you already know the 3D structure? Can you go through each residue and (easily) figure out what the effect of replacing each of the other 19 amino acids at that position would be? It seems like you could...just load the structure into Rosetta or similar software, replace the amino acid, let it wiggle around a bit. If the new acid doesn't physically fit, the energy goes way up, or the RMSD changes too much, that amino acid is not a good candidate to be replaced in that position.

After you went through the entire known structure, you'd have a "regular expression" that you could use to check unknown sequences to predict if they have similar folds, even if the raw sequence wasn't so close.

---

But what if you already know the 3D structure? Can you go through each residue and (easily) figure out what the effect of replacing each of the other 19 amino acids at that position would be? It seems like you could...just load the structure into Rosetta or similar software, replace the amino acid, let it wiggle around a bit. If the new acid doesn't physically fit, the energy goes way up, or the RMSD changes too much, that amino acid is not a good candidate to be replaced in that position.

Yes, you are right, just what came to my mind after I sent the post: you were talking of variations on a known structure. That would make sense. It is kind of the "in silico" variant of side chain changes in pharmaceutical research, for example in search of new chemotherapeutics.

You would of course still have to test it in the docking computation to make sure it is as good as the original. I am not sure how far you have read through the sections given on the Rosetta Home Page, there is a lot said in that regard, especially concerning Protein Design.

Rosetta already does a lot of what you have in mind, although till now the aim is rather to synthesize Proteins with a desired functionality from scratch or intentionally altering known Proteins to adopt new functionality. [On the other hand, thinking of legal consequences, if a protein is patended, the question is if the functionality itself is covered by that patent even if you alter the sequence a little. It would be like rebuilding a patented device with slightly altered components and claiming it was something generically new. It may not be easy to circumvent a patent that way.]

In general, in order for something to be patentable, it has to be a less then obvious change from existing published or patented items. But that simply speaks to the altered, but similar, protein and implies that the new protein is not patentable, because it is a simple, obvious (to a protenomics researcher anyway) change to make from the existing well-known protein.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/