Detailed description of the project and the tools we use
General Description of the project
The Go Fight Against Malaria project uses AutoDock 4.2 and the new “AutoDock Vina” computer software to evaluate millions of candidate compounds against at least 15 different molecular drug targets from the malaria parasite in order to discover new inhibitors that can block the activity of these multi-drug-resistant mutant superbugs. The first half of the project will search for these promising candidate compounds with Vina, and the second half of the project will use AutoDock 4.2. These compounds will be tested by "docking" flexible models of them against 3-D, atomic-scale models of different protein drug targets from the malaria parasite, to predict (a) how tightly these compounds might be able to bind, (b) where these compounds prefer to bind on the molecular target, and (c) what specific interactions are formed between the candidate and the drug target. These calculations will be used to predict the affinity/potency of the compound, the location where it binds on the protein molecule, and the "binding mode" it uses to potentially disable the target. Compounds that can bind tightly to the right regions of particular proteins from the malaria parasite have the potential to “gum up” the parasite’s molecular machinery and, thus, help advance the discovery of new types of drugs to cure malaria. Since these predictions are not perfectly accurate, the top-ranked candidate compounds we discover in these virtual experiments will then be tested in “biological assays” performed by our collaborators in test tubes and Petri dishes.
Once our collaborators have proven that some of these candidate compounds are definitely able to help kill the malaria parasite, then The Scripps Research Institute and other researchers throughout the world can try to optimize these promising compounds to increase their potency against the target while decreasing their ability to bind to human proteins (since unwanted binding to human proteins can cause toxic side effects). Once it is known that a compound is a novel inhibitor of one of these drug targets, scientists called "medicinal chemists" can then extend and modify these compounds in order to accelerate the development of new anti-malaria drugs.
General description of the "docking" calculations we are performing
We computationally predict how potent these candidate compounds might be at disabling key target molecules from the malaria parasite by performing “docking” calculations. Docking is a way of trying to figure out how well a small chemical compound can bind to and block the activity of a target protein. These calculations predict how potent the small chemical compound might be at blocking the activity of the target molecule, the location where it binds on the protein target, and the detailed binding mode it uses to potentially disable that target. It's like trying to find the right key to open a particular lock. However, both the lock and the keys are flexible—they can change shape, or transform their conformation, as they wiggle, jiggle, dance, expand, and contract in the warm watery environment in which they reside. In addition, some locks have multiple different types of keyholes, and only one or two of them might be useful at disabling the parasite's molecular machinery. When, by chance, the lock's internal structure happens to change a bit, the potential for evolutionary improvement occurs. If that change in the lock's guts happens to help the parasite escape the effects of a key/drug, then that new lock becomes a “drug-resistant mutant”. To make it even trickier, the total number of potential keys that could exist in the universe (the size of “chemical space”) is estimated to be about 10 to the 60th power (that is, 1 with 60 zeros after it). We obviously can't computationally evaluate flexible models of that many different keys, so we'll focus these experiments on the types of keys that are somewhat similar to the types of molecules that have already become approved drugs. In the GO Fight Against Malaria project, we will computationally evaluate millions of chemical compounds (potential keys) against models of at least 15 different drug targets from the malaria parasite. Compounds that can bind tightly to the right regions of particular proteins from Plasmodium falciparum have the potential to “gum up” the parasite's machinery and, thus, help advance the discovery of new types of drugs to cure superbugs of malaria.
Detailed, technical description of the "docking" calculations we use
This project will use two different types of “docking” programs to search for new compounds that can bind to and block the activity of protein drug targets from the malaria parasite. Both of these docking programs were created and developed by the Olson lab at The Scripps Research Institute (http://mgl.scripps.edu). The first phase of the project will computationally evaluate the potential potency of millions of compounds using the new software “AutoDock Vina,” which was created by Oleg Trott and Arthur J. Olson. The second phase of the project will computationally re-evaluate the potency of the same compounds using the program “AutoDock4.2,” which was created by Garrett M. Morris, Ruth Huey, William E. Hart, William Lindstrom, Alexander Gillet, David S. Goodsell, and Arthur J. Olson. These two different types of docking programs each use different algorithms when searching for the location where a compound binds and when predicting the detailed mode it uses to bind to that location of the protein target, and they both use different “scoring functions” to evaluate the potency of the binding mode they predicted. Since no computational tools are perfectly accurate, harvesting compounds that score well with multiple different types of computational tools can increase the probability of discovering promising new compounds. In the Olson lab’s experience with the FightAIDS@Home project (see Volume 10 at http://fightaidsathome.scripps.edu), evaluating compounds with both AutoDock and Vina facilitated the discovery of novel inhibitors of HIV protease (which is a notoriously difficult protein to target).
AutoDock is a suite of automated docking tools designed to predict how “small molecule compounds,” such as substrates or drug candidates, bind to a receptor (target) of known 3-D structure and to estimate how tightly that small molecule binds to that receptor. That is, it predicts both the small molecules “binding mode” and estimates its potency against the target. The AutoDock algorithm is essentially a high dimensional stochastic search utilizing a “Lamarckian genetic algorithm” approach with flexible models of the small molecules. When docking any given drug candidate against a particular protein target, the space of all possible configurations must be explored to find the best energetic fit between the two molecules. Any given docking protocol must explore all possible degrees of freedom that are specified in the system. AutoDock consists of two main programs: autodock performs the docking of each compound against a set of grid maps that describe the energetic landscape of the target protein, and autogrid pre-calculates these grid maps before autodock is ran, which greatly increases the speed of the autodock phase of these calculations.
AutoDock Vina also uses pre-calculated grid maps (which are generated internally, instead of using a separate program, such as autogrid). Vina also uses flexible models of the small molecules, and it also treats the docking process as a stochastic global optimization of the scoring function. But Vina utilizes a different scoring function and a different search algorithm than AutoDock, and Vina’s search process is guided by the gradients in the energetic landscape of the target protein (unlike AutoDock4.2). However, Vina uses the same “pdbqt” file format that AutoDock4.2 uses to represent the atomic models of both the target protein and the small molecules, which makes the process of preparing and analyzing the Virtual Screening experiments for GO Fight Against Malaria that will use both docking programs much, much easier.
The naming convention for the GO Fight Against Malaria work units
If you click on the "advanced view" of BOINC for World Community Grid, the names of the jobs we send out start with the abbreviation for the project (GFAM), followed by my internal filename for the target, followed by a unique batch identification number, followed by a work unit identification number. Each batch involves docking 10,000 different compounds with Vina against a single target. The batch numbers are described on the page with "Experiments we're performing and our future plans." The names of the targets start with "x" to indicate that they were superimposed onto a specific, single coordinate reference frame. The x is generally followed by the PDB ID code for the structure of the target being used (see the Protein Data Bank at http://www.rcsb.org/pdb). NDP refers to the cofactor that is bound to a region of the active site in dihydrofolate reductase (DHFR, the first target for this project). If the target has WATs in the name, then some of the waters from the crystallographic structure of the target were included. In "positive control" docking calculations that involve reproducing the known binding mode of current inhibitors of DHFR, these crystallographic waters can sometimes improve the accuracy of the calculations. Other targets have "dry" in the name, to indicate that all of the water molecules from the crystallographic structure of the target were removed. If a compound can displace these crystallographic waters (which we can help determine by docking against both wet and dry versions of the target), then it might receive an extra energetic boost during the binding process.
Most of the targets are from Plasmodium falciparum, the species of parasite that causes the deadliest form of malaria, but some of the jobs involve the same type of target from Plasmodium vivax, the species of malaria that causes the most malaria infections (but vivax infections tend to be less fatal than infections with falciparum). Finding new compounds that can inhibit the targets from both Plasmodium falciparum and Plasmodium vivax would be very clinically useful.
Some targets have TB in the name, which indicates that these targets are enzymes from Mycobacterium tuberculosis. The three deadliest infectious diseases are HIV, malaria, and tuberculosis. Some of these GO Fight Against Malaria experiments involve screening compounds against the versions of a particular drug target from both malaria and tuberculosis; thus, we'll be sneaking in some tuberculosis research while we fight against malaria. This is scientifically justified, since it is known that some potent inhibitors of dihydrofolate reductase (DHFR) from malaria are also able to inhibit DHFR from tuberculosis. Similarly, some of the inhibitors of malaria's enoyl-acyl-carrier-protein reductase (ENR) are also good inhibitors of the version of ENR from tuberculosis (which is called "InhA"). Consequently, searching for compounds that can inhibit these targets from both malaria and tuberculosis could be very useful against malaria, and it could help advance the research against two of the deadliest infectious diseases on the planet.
The two images above show the results of a "positive control" experiment with Vina against Pf ENR (enoyl-acyl-carrier-protein reductase). The image on the right is a close-up of the active site region (the specific keyhole we are trying to disable), while the full target is displayed on the left. These images demonstrate that we can accurately predict/reproduce the detailed "binding mode" that a known inhibitor uses to disable this key drug target. The known, experimentally-determined, X-ray crystallographic binding mode of this inhibitor is shown in magenta, while the results from the Vina calculations are displayed in cyan.
A few experiments also involve the version of a particular enzyme from humans (such as DHFR). We want to find compounds that are predicted to inhibit the targets from malaria, but we don't want these compounds to strongly inhibit the enzymes in humans (since inhibiting certain human enzymes can cause toxic side effects). Thus, to advance research against malaria, we'll harvest compounds that dock well against the targets from malaria and that also score poorly against the human version of that enzyme. However, since all data from this project are in the public domain, the data on the compounds that happen to score very well against the human DHFR could help other scientists advance the search for new drugs against some types of cancer or new drugs to prevent the rejection of transplanted organs. I like to design experiments that can produce multiple bangs for the buck. ;)
The molecular targets at which we are aiming
Both AutoDock Vina and AutoDock4.2 will be used to screen millions of candidate compounds against at least 15 different “validated drug targets” and “potential drug targets” from the malaria parasite. Basically, these experiments will target every relevant protein from the malaria parasite that has an atomically-detailed 3-D structure available. GO Fight Against Malaria will screen candidate compounds against the following protein targets from the malaria parasite: dihydrofolate reductase, enoyl-acyl-carrier-protein reductase (also known as Fab I), purine phosphoribosyltransferase, purine nucleotide phosphorylase, M1 neutral aminopeptidase, falcipain (a cysteine protease), glutathione reductase, glutathione S-transferase, dihydroorotate dehydrogenase, orotidine 5’-phosphate decarboxylase, merozoite surface protein-1, profilin, 3-oxoacyl acyl-carrier-protein reductase (also known as Fab G), and beta-hydroxyacyl-acyl-carrier-protein dehydrase (also known as Fab A/Z).
All “GO Fight Against Malaria” results will be in the public domain in the form of the virtual screening data that will be generated on World Community Grid and will be freely available to the global community of malaria researchers. Consequently, many other labs throughout the world will be able to use these results to help them discover new anti-malaria compounds that they and The Scripps Research Institute can then develop into new classes of drugs to treat this severe and neglected disease.