Designing Proteins In Silico

Fluorophore writes "In a recent issue of the scientific journal Nature, scientists in the lab of Homme Hellinga at Duke University reported designing proteins using a cluster of 20 computers. These proteins were then tested in the lab and shown to bind their intended targets including TNT, serotonin and lactate. This is a tremendous step for computational biology, nicely reviewed in C&E News' top story. Designer proteins such as this can be developed for bioremediation of weapons dump sites (TNT) and sensitive sensors of drugs/contaminants that can easily be grown in bacteria."

Comments

More Info

[Bowling Moses]

The actual Nature article is "Computational design of receptor and sensor proteins with novel functions," in the May 23, 2003 issue (Vol 423 No 6936 pp101-205). It is important to note that they are not making fully functional enzymes yet, but have accomplished the rather daunting task of designing/directing the evolution of a given protein binding substrate A and making it bind a new, completely different substrate B. Their wild-type substrate interacts with 12-18 residues, so multiply that by your 20 standard amino acids across these interacting residues and you have a crapload of sequences to deal with (10^15 to 10^23; I'll take their word for it). I thought the statement "Designer proteins such as this can be developed for bioremediation of weapons dump sites (TNT) and sensitive sensors of drugs/contaminants that can easily be grown in bacteria." was kind of cute as when you search Pubmed with "TNT reductase" you get back a number of articles on bacterial enzymes that allow them to munch TNT. A few years back I got to work on a project to solve the structures of enzymes that pop NO2 groups off of TNT and related compounds; the bacteria that these proteins were subcloned out of were found in the heavily contaminated soil of a former World War 2 munitions plant. Pretty cool what evolution can do when you add a new component to the environment of some organism.

Re:More Info

[Fluorophore]

Just a tiny factual correction to the above post. For those searching for this article, its in the May 8th issue of Nature. The remainder of the above reference is accurate.

What makes this success deserve the superlative 'humongous', imho, is twofold. One, as Bowling Moses refers, the size of sequence space is 10^15 to 10^23. However, you combine this with the number of possible rotameric conformations certain sidechains can adopt, and your search space climbs to 10^50 to 10^70 in size. Make things even more formidable by thinking about the rotational and translational degrees of freedom within an active site pocket and you're trying to find the best solution from among 10^110 states!! Only because of novel improvements to Dead End Elimination, which were outlined in an earlier article by Looger and Hellinga in the Journal of Molecular Biology, are such huge problems able to be solved in 3 days. The second major triumph of this paper is the design of polar specificity. While not the first example of designing polar interactions, a strict rule of satisfying all possible hydrogen bonds has greatly improved both Dead End Elimination selection and the specificity of the resulting active site. Up until now, the best designed small molecule binders were shape matching grease with grease.

When using a higher resolution search scheme (e.g. more states), Looger and colleagues were able to design a TNT binding protein with nanomolar binding, while preserving specificity. It seems possible that Hellinga may be at the top of the pack in designing useful enzymes. If the proteins are designed to target a substrate transition state, it may be possible to design artificial biocatalysts.

While plenty of TNT degrading enzymes have been developed either by natural evolution or by artificially directed processes, the advantages of an in silico approach are obvious. Hellinga could make one design in 3 days ... and computers will only get faster.