RESEARCH HIGHLIGHTS

CHEMICAL BIOLOGY From structure to function methods By docking potential substrates into the thioadenosine, S-adenosyl-L-homocys- of an of known struc- teine (SAH) and adenosine itself, were bona ture, researchers accurately predict the fide substrates of the enzyme. Notably, the catalytic activity of the enzyme. researchers determined the X-ray crystal .com/nature e In recent years, structural genomics ini- structure of Tm0936 with the of tiatives have had great success in generat- the SAH deamination reaction, and found .natur

w ing large numbers of structures of so-far that the docking prediction and the actual uncharacterized proteins. But, as Brian structure of the complex had close congru- Shoichet at the University of California at ence (Fig. 1). Thus, Tm0936 was identified San Francisco puts it, “even if you know what as an . http://ww

a protein looks like, this doesn’t necessarily In another recent related study,

oup mean you know what it does. We decided to docking was applied to predict function of a r

G take the next step and ask [whether] we can member of the enolase superfamily. In that

broadly predict function of an enzyme if we Figure 1 | Comparing predicted with actual case, a homology model of the test enzyme, know the structure.” In collaboration with structure. The high-energy form of SAH (green) rather than a crystal structure, was used as lishing Frank Raushel at Texas A&M University the template for docking (Song et al., 2007). b in the active site of Tm0936, as predicted and Steve Almo at Albert Einstein College by docking, was superimposed on the crystal Subsequent experimental studies led to Pu of Medicine, Shoichet and his group have structure of the enzyme-substrate complex identification of the enzyme as an N-suc- tackled this question for Tm0936, a protein (substrate in red). Reprinted from Nature. cinyl arginine/ racemase. The amido-

Nature from the microorganism Thermotoga mari- and enolase superfamilies contain

7 tima (Hermann et al, 2007). of the test molecules. As the enzyme under about 6,000 and 1,000 proteins, respectively.

200 The researchers took a computational study, Tm0936, is broadly classifiable as The stage now seems set for function pre-

© docking approach, in which a database of belonging to the amidohydrolase superfam- diction for more of these. potential substrates is docked into the active ily, the researchers could limit the possible There are certainly challenges ahead, site of the enzyme. The degree of structural forms of the test substrates by conversion however. The right substrate for a particular complementation, or fit, of a given substrate only of the appropriate functional groups. enzyme may simply not be in the database could then predict what the activity of the Beginning with a good database is critical, being tested. In addition, problems could enzyme is. This approach has been used and creating one is not easy. “If you’re talk- arise for that undergo substantial before for the design of inhibitors, but the ing about 4,000 molecules with 20,000 states, conformational change during catalysis. challenges for substrate prediction are much that’s 20,000 problems,” Shoichet says. “Each Although retrospective studies indicate that greater. “Inhibitor design is hard enough,” molecule has its own questions‘What’s docking high-energy intermediates makes Shoichet says, “but in that case all you have the protonation state?’, ‘What’s the isom- this sort of predictive approach less suscep- to do is get a molecule that the enzyme rec- erization state?’, ‘Do you have to test all the tible to conformation-based problems, it ognizes. It doesn’t have to do the next step, possible stereocenters?’, ‘Which ones are does not eliminate them completely. which is to turn the molecule over, to do the likely to be the most biologically relevant?’.” But the structural genomics pipelines catalytic reaction on it. And capturing that All the more laudable, therefore, is that the continue to flow, the raw material is at is no joke.” results of generating the test structures and hand, and the extent to which the docking To address this challenge, the researchers docking them into the active site of Tm0936 approach is successful at going from form to docked not the ground-state of substrates were so informative. function will become clear as it is applied to but rather a high-energy intermediate state When the substrates were ranked in terms more and more protein structures. that would be recognized by the enzyme, if of fit, the researchers found adenosine Natalie de Souza catalysis were possible with that substrate. and several adenosine analogs among the RESEARCH PAPERS For the 4,207-metabolite database that highest ranked molecules; nine of the ten Hermann, J.C. et al. Structure-based activity they screened, this involved computational top-scoring hits belonged to this category. prediction for an enzyme of unknown function. conversion of all of the relevant functional They tested four of these molecules experi- Nature 448, 775–779 (2007). Song, L. et al. Prediction and assignment of function groups to their high-energy states, and mentally for catalytic turnover by Tm0936, for a divergent N-succinylamino acid racemase. Nat. resulted in a final set of about 22,500 forms and verified that three of them, 5-methyl- Chem. Biol. 3, 486–491 (2007).

NATURE METHODS | VOL.4 NO.10 | OCTOBER 2007 | 771