Tetrazole scaffolds

Tetrazole is a common drug scaffold. It's found, for example, in Losartan, a bi-phenyl tetrazole based drug and angiotensin II receptor antagonist. The figure shows a tetrazole scaffold and electrostatic potential maps:


Tetrazole scaffolds

Three of the four N-atoms on tetrazole are clearly nucleophiles because of the presence of lone pair electrons. The fourth N-atom is apparently shielded by the bonded H-atom. The tetrazole heterocycle should bind strongly to positively charged binding sites, which is in fact what happens in angiotensin receptors. The electrostatic maps show fairly dramatically why this binding should occur.

What the doctor ordered

One of my "hobbies" is to investigate the molecular structure and physicochemical properties of drugs that are prescribed by my doctor. (When you start doing this you know you've been in the field too long).

I've been doing too much computer work lately so I have a sore back. My doctor prescribed cyclobenzaprine (Flexeril) as a muscle relaxant. Here's the 2D molecular structure:


Cyclobenzaprine

I was curious to know if this drug followed Lipinski's rules for "drug-like" compounds. Here's some data:

Molecular weight: 275
H-bond donors: 0
H-bond acceptors: 1
Rotatable bonds: 3
LogP: 4.9

These values do indeed fall within the range of Lipinski's rules. Also, the drug really works - my back feels much better now.

Leads and drugs: what's the difference?

A client once asked us: "What's the difference between a lead compound and a drug compound?". It's an excellent question. The answer is not cast in stone, but Oprea et al. (J. Chem. Inf. Comput. Sci. 41:1308-1315, 2001) addressed the issue.

They found that drug compounds were generally more complex (higher molecular weights, more ring structures, more rotatable bonds), more hydrophobic, and had more hydrogen bond acceptors.

Lead compounds are often derived from virtual high-throughput screens. Based on Oprea's results we should expect these compounds to become more "complex", hydrophobic, and involve more H-bonds as they are optimized into drug compounds.

It's probably worthwhile keeping this study in mind during lead optimization research. Here's a figure that shows some leads and the drugs that eventually evolved from them.


Optimization of Ranitidine, Mifepristone, and Diltiazem.

How much sequence is enough?

We're often asked to build 3D homology models of proteins based only on protein amino acid sequence. The basic idea is to predict a protein's 3D structure (target) by comparing its AA sequence to sequences from similar proteins with known 3D structures (templates). The perennial question that comes up is "How similar do the target:template sequences have to be for you to trust the 3D models?"

Chakravarty et al. (Nucleic Acids Researsh, 33(1):244-259, 2005) have some interesting results on this point. They show that 40% sequence identity is probably good enough for many homology models. In fact, with 40% identity, the homology models are about as accurate as NMR structures, and NMR structures are usually considered good enough for modeling drug interactions with proteins. (However, the 3D structures from homology models are not as good as those derived from X-ray crystallography, which are considered the gold standard).

This figure shows the 3D X-ray structure for a Staphylococcus protease and several homology models. The match is pretty good at 60% and 30% sequence identity.


Surface and backbone structures of S. griseus protease B. Left-hand shows X-ray structure; remaining four show homology models. Sequence identities for the models are 61%, 32%, 22%, and 13%.

The Bioinformatics Channel

Today we launched The Bioinformatics Channel
(http://www.b-eye-network.com/channels/index.php?filter_channel=1131&PHPSESSID=32157b48ea7addbb9e2584323900b2ed).

The Channel includes monthly articles on current topics in bioinformatics, industry events, news releases, insightful opinion pieces, industry and trade association links, white papers and reports, and other topics of interest to the Bioinformatics community. We hope you check out the site.

New release from OpenEye

We make extensive use of CADD software from OpenEye Scientific Software (www.eyesopen.com) in our drug discovery and design services. OpenEye recently issued a beta release of Vida 2, their graphical interface to their software suite. The final release is due out later this year. A new ribbon display mode is available in Vida 2. Here's a screenshot of p53 (PDB ID = 1TSR) complexed with DNA:



p53 ribbon view

Alpha helices are shown in red, beta sheets in yellow, and loops in grey. This capability will give new insights on protein structure.

Protein Structure Validation

Homology modeling is one of the most widely used methods for modeling the 3D structure of proteins. Starting with only amino acid sequence data, we can model the 3D structure of the sequence (the target) by comparing it with known 3D structures of related proteins (the template). However, before running a homology model it's important to know how "good" the template structure is.

One popular method is to use the ANOLEA (atomic non-local environment assessment) bioinformatics tool. ANOLEA examines the interactions between heavy atoms in amino acid residues, computes the mean force potential between them, and generates an energy profile for the 3D structure of a protein.

Here's an example of an ANOLEA energy profile for the PDB (www.pdb.org) file 1OPA, retinol binding protein II:


ANOLEA Energy Profile for 1OPA

All the residues are "in the green" (except for one or two that are just slightly "in the red"). The energy profile suggests that there are no high energy amino acid residues. 1OPA would probably be a good template structure in homology modeling. We've used ANOLEA in numerous homology modeling projects and found it to be very useful in validating protein structure.

Smallest Protein

Ever wonder what the smallest protein is? Apparently it's TRP-Cage, a protein with only 20 amino acids derived from the saliva of Gila monsters.


Trp-cage - smallest protein

You can find the structure file and images in the PDB database (www.pdb.org) with PDB ID = 1L2Y. This highly stable mini-protein is important for studies of protein stability, protein folding, and 3D structure. Even with this small size, it displays secondary structural elements, such as an alpha helix, found in many proteins. So far there are no known proteins with less than 20 residues, but we'll see what happens in the future.

Bioinformatics Software

Bioinformatics software - not for the faint of heart.

While compiling a list of bioinformatics software applications and databases that we use for CADD research, I discovered that we have 100+ packages and databases in our portfolio (!). It covers things like sequence alignment, homology modeling, molecular modeling, virtual high-throughput screening, geometry optimization, energy minimization, quantum mechanics - the list goes on and on. And it will continue to rise as we tackle more client projects.

Some of these are commercial packages, many OpenSource, and a few home-grown. Clearly, it's a challenge trying to master all of them and stay current with updates. It also demonstrates the need for Bioinformatics training in the US. I hope the universities respond with training and programs in Bioinformatics.

Solvation models.

Protein conformations are often modeled in vacuo, that is, without including the normal aqueous environment found in vivo. However, accurate physiological representations of protein 3-D structure should probably include water molecules and hydration shells. The importance of this is driven home in the article "Computer-aided development and use of three-dimensional pharmacophore models" by Liljefors and Pettersson, in the Textbook of Drug Design and Discovery by Krogsgaard-Larsen et al., see page 98. The figure is reproduced below.

The 2-D structure of biotin is shown at top. The preferred gas phase 3D structure is shown in middle left and the preferred solution phase 3D structure is shown in middle right. The in vacuo structure is pseudocyclic with an H-bond between the carboxylate and amine groups; the hydrated structure is linear with no H-bond formation. The gas phase and solvated structures are completely different. Furthermore, when biotin binds with streptavidin it maintains the linear shape, suggesting that the solvation is critically important for proper binding (bottom figure).

Even though it's computationally more expensive to run solvated models, we should always consider this when modeling protein-ligand interactions.



Solvation model of protein conformation.

BioWest Conference

We'll be attending the annual BioWest Conference in Denver, CO (http://www.biowestconference.com/) in a few days. If you're in the Denver area on Oct. 26, 27, stop by and check it out. Great event for bioscience enthusiasts.

BlueGene tops supercomputing list

IBM's BlueGene supercomputer soared to number 1 on the list of the world's fastest supercomputers with a record breaking performance of 36 Tflops (http://news.com.com, search for "Blue Gene"). It narrowly edged out Japan's Earth Simulator supercomputer, which comes in at 35 Tflops. These marks are based on parallel Linpack benchmarks, which favor dense linear systems of equations (see www.top500.org). Although BlueGene is not yet officially the fastest supercomputer, as recognized by the Top500 list, it will undoubtedly be there when the next list comes out in November, 2004.

This is an important event for CADD research (computer aided drug design) as BlueGene was specifically designed for the Life Sciences market. BlueGene can be used to simulate biomolecular phenomena, such as protein folding and protein-ligand docking, both of which are crucial in understanding the physicochemical properties of drug-receptor interactions. Computational modeling of drug compounds is often limited by available compute power. BlueGene will push the envelope in supercomputing and hopefully provide the horsepower needed to improve CADD models of new drug compounds.

Drug discovery costs

One of the hot topics in our political environment today is the cost of consumer drugs. Much heated debate surrounds the high costs for drugs, especially in the US. There are two reports that address this issue - the Tuft's Report and the PhRMA Industry Report (http://www.rmcsoft.com/docs.cfm). The average cost for drug development is $500M - $1.5 B and 7 - 15 years of research and development effort to bring one drug to market. Computer aided drug design methods may help bring down the cost of drug development by guiding and directing experimental research towards the most promising drug candidates. Using CADD methods to reduce drug development costs by just 5 - 10% would be a significant contribution, and could help lower the political temperature around this issue.

Reductionist thinking

For those of us who work with molecular models every day, we often find ourselves applying reductionist thinking in drug design research. That is, we tend to focus on the physicochemical behaviour of drug compounds docking with receptor targets. This is often viewed as an atomic scale phenomenon. Hence, we take the reductionist view of drug discovery and design as a series of docking events between small-molecule compounds and their targets.

However, two recent provocative and intriguing articles argue for a more functional, integrative, systems biology point of view (see Walker et al., "Functional pharmacology: the drug discovery bottleneck?", Drug Discovery Today, vol. 3(5), pp. 208-215, 2004; Buehler, "Advancing drug discovery - beyond design", PharmaGenomics, July/August, pp. 24 - 26, 2004). Both authors suggest that drug design research protocols should focus on hierarchical levels above that of simple protein-ligand docking. They suggest viewing drug design in terms of cell signaling pathways, metabolic pathways, molecular ensembles, and in-vivo pharmacology. These higher integrative levels may more accurately describe the actions of drug compounds.

Perhaps reductionist and integrative approaches should be pursued simultaneously. Recent trends suggest that virtual high-throughput screening and docking studies should be immediately followed by bioavailability and bioactivity studies (i.e. reductionist -> functional) and then back again if necessary. This iterative approach combines the best of both worlds.

SpaceShipOne

Congratulations SpaceShipOne!

Burt Rutan and Scaled Composites just won the $10M Ansari X Prize for privately funded sub-orbital space flight. What a great achievement. We congratulate them for this wonderful feat of engineering and personal human endeavor.

Virtual Screening

We found an interesting presentation, "Large-Scale Virtual Screening for Novel Inhibitors", which is available at http://www.rmcsoft.com/docs.cfm. The author screened a 2.7 M compound library using FlexX in a search for urease inhibitors. On a parallel Sun Fire cluster (40 CPU's) the screen required 26 days for completion., ca. 33 sec. per docking event. Several promising lead compounds were identified. This is a good example of the vHTS research protocol.

Pharma IT/IM

Information technology and information management are central elements in virtually every pharmaceutical and biotechnology company engaged in drug discovery. And the successful management of IT/IM projects is critically important for these companies. The Project Management Institute (www.pmi.org) runs a Pharmaceutical Special Interest Group (www.pharmasig.org) that specializes in project management for this sector. You might check it out if you're in any way involved in PM activities for a pharma/biotech company.

Blue Gene goes commercial

According to Bio-IT World ( Sept. 7, 2004, http://www.bioitworld.com/news/090704_report5977.html), IBM's Blue Gene/L supercomputer has found its first commercial customer. Apparently, the Computational Biology Research Center in Japan purchased a Blue Gene system for conducting research in protein folding and drug design. With peak processing power of about 22 Tflops it should offer a substantial improvement in the CBRC's ability to simulate the dynamics of protein folding. It's encouraging to see commercial activity in both high-performance computing and protein folding.

High-performance computing article

In Drug Discovery Today - Biosilico, Sept. 2004, v.2(5), p. 175, there is an informative article on the use of high-performance computing systems in drug discovery. Most notable is the continuing steady rise in cluster computing, where largely commodity parts are used to build powerful clustered systems. Apparently more than one-half of the Top500 (www.top500.org) supercomputing sites are now cluster systems, largely displacing symmetric multiprocessors (SMP's). The article lists hardware and software vendors that service the cluster computing market. It's well worth a read, especially if you're designing computational infrastructures for drug discovery.

ChemInfo conference

Don't miss the upcoming virtual conference Applications of Cheminformatics and Modelling to Drug Discovery. They've booked an exemplary roster of speakers for the conference.

I'm especially interested in the seminars on QM/MM (quantum mechanics / molecular mechanics) hybrid methods. There is some question about how well the interpolation schemes handle QM/MM interfacial atoms, and the potential introduction of numerical errors from link atoms. Let's see if anyone addresses this in the seminars.

Online lectures

If you're interested in reading excellent online lecture material about the electronic structure of atoms and macromolecules, check out Dr. Richard Bader's web site. The material presented there is directly applicable to drug-receptor binding properties. It's eminently readable and includes many valuable diagrams that help explain the concepts. I liked it so much that I printed it to real paper.