|By: Dr. Alexander Luke Perryman|
|Research Teaching Specialist III|
|14 Jul 2014|
Dr. Alexander Perryman describes the analysis and initial findings from the first phase of GO Fight Against Malaria, which include the discovery of several promising hits against key drug targets for treating both malaria and drug-resistant strains of tuberculosis. They are conducting further analysis and experimentation on the massive amount of data generated by World Community Grid volunteers.
Dear fellow volunteers of World Community Grid,
In under two years, World Community Grid volunteers performed the world's largest docking project, carrying out over 1 billion calculations to help us identify chemical compounds to advance the treatment of increasingly drug-resistant strains of malaria and other diseases - a process that would have taken over a hundred years on the type of computer clusters currently available at most universities.
Since we completed GO Fight Against Malaria (GFAM) calculations on World Community Grid a year ago, we've been analyzing the generated data. Although that process will continue for some time still, early analysis has revealed several promising findings.
First, we identified the first "small molecule" inhibitor (i.e., drug-like compound) to block the activity of a particular malaria enzyme involved in infection, the first step in developing a potential treatment or prevention aimed at this malaria drug target.
Also, a subset of your calculations was conducted against a drug target for malaria which shares a similar atomic structure to a Mycobacterium tuberculosis enzyme. With extensively drug-resistant strains of tuberculosis on the rise, there is a pressing need to identify more effective treatments. We therefore included this particular tuberculosis drug target in our GFAM experiments. In doing so, we have identified several chemical compounds as potential inhibitors of this enzyme and have confirmed these results with initial laboratory tests. A very impressive number of the promising chemical compounds identified through the virtual screenings you computed on World Community Grid have gone on to perform well in additional lab testing: 20% were "hits", vs. less than 1% on average for other experimental ("wet lab") high-throughput tuberculosis experiments.
We are now designing and synthesizing new derivatives of these inhibitors to further refine them as viable drug candidates. Read on for more details about this early analysis work, and we'll be able to share more information once we publish our findings. In the meantime, I want to thank GFAM volunteers for allowing us to advance this important and often neglected area of research.
Largest set of computational docking experiments ever performed
GFAM was launched on IBM's World Community Grid on November 16, 2011. Malaria is one of the three deadliest infectious diseases on Earth (the other two are HIV and tuberculosis). Plasmodium falciparum (Pfal, or Pf), the species that causes the worst form of malaria, kills more people than any other parasite on the planet. Over 200,000,000 clinical cases of malaria occur each year, and over one million people are killed by malaria every year. Over three billion people (almost half of all humans) are at risk of becoming infected with malaria, and every 30 seconds, another child dies of malaria.
GFAM ran on World Community Grid for 19 months, during which the tremendous computational power provided by World Community Grid volunteers like you helped us generate massive data sets against 22 different types of drug targets, to seed the discovery of new drugs to treat malaria. We performed "docking calculations," which explore how well different "small molecules" (pieces of drug-like compounds) are able to bind and potentially block the activity of critical pieces of the molecular machinery that the pathogens use to survive, replicate, and spread throughout humanity. Docking calculations use flexible models of these small molecules to explore the energetic landscape of atomic-scale models (on the scale of 0.0000000001 meters) of proteins that perform critical functions for the parasite's lifecycle and infection process. These calculations predict how tightly a compound might bind to the target (that is, how potent it might be), where the compound probably prefers to bind, and what specific types of interactions might be formed between the compound and the drug target. One docking calculation refers to the process of docking a flexible model of a single compound against one particular version of one target. In this first phase of GFAM, World Community Grid volunteers performed 1.16 billion different docking calculations that explored the potential activity of 5.6 million different compounds against drug targets from malaria (and against some targets for treating drug-resistant tuberculosis, Methicillin-Resistant Staphylococcus aureus (MRSA), filariasis, and bubonic plague, when the targets from those other pathogens had structural similarity to the targets from malaria). With the computing power that you generously donated, GFAM was the first project to ever perform a billion different docking calculations. Performing this many calculations could have taken over a hundred years on the type of computer clusters currently available at most universities. We could not have accomplished this feat without your help. We are also grateful for the $50,000 in seed funding provided by the IBM International Foundation, from part of the prize money that IBM's computer Watson won on Jeopardy!. Thus far, that seed money has been the only funding that the project has received, but we are currently writing grants that focus on analyzing and extending the GFAM data.
Finding a "hit" that inhibits a critical protein involved in malaria infections
"Hits" are compounds that have some inhibitory effect on the biological activity of one of these drug targets. But finding a hit is only the beginning of the process (a complicated process that can take several years to a couple of decades to complete). Scientists from around the world called "medicinal chemists" can then work with structure-based computational chemists like us to try to increase the potency and decrease the potential toxic side effects of these compounds, which involves processes called "hit-to-lead development" and then "lead optimization". "Leads" are larger, more structurally complex, potential drug candidates that generally display nanomolar potency (that is, they are around 1,000 times more potent than "hits", which means that only a tiny amount of a leading compound is required to affect the activity of the target). In collaboration with Professor Mike Blackman's lab in the Division of Parasitology at the Medical Research Council's (or "MRC's") National Institute for Medical Research (or "NIMR"), in London, UK, and with InhibOx, Ltd, we searched for the first small molecule inhibitors of the potential drug target "PfSUB1" (see target class #6 on http://gofightagainstmalaria.scripps.edu/index.php/how-we-will-discover-potential-malaria-drugs). When the Blackman lab solved the first crystal structure of PfSUB1 (that is, the atomically-detailed, 3-D map of where all of its atoms are), they shared that unpublished structure with us, which allowed us to perform virtual screens against PfSUB1. These virtual screens are searching for "small molecule" inhibitors (that is, compounds with some similarity to pieces of known drugs) that can block the activity of this malarial enzyme. When malaria parasites replicate themselves inside a red blood cell, the "daughter" parasites eventually rupture the infected host cell, which allows the new parasites to escape and then invade and infect other red blood cells. The subtilisin-like serine proteases from Plasmodium falciparum (also known as PfSUB1) are involved in this ability of the malaria parasites to escape (or "egress") an infected red blood cell. The Blackman lab has shown that the PfSUB1 enzyme has an additional role in "priming" the merozoite stage of the parasite prior to its invasion of red blood cells (in other words, it is involved in processing certain other malarial proteins in order to prepare and activate them, so that the parasite can invade our blood cells). Thus, PfSUB1 is involved in both the egress and the infection process. In the results of GFAM Experiment 27, we discovered the first small molecule inhibitor of PfSUB1 ever identified, and it displayed a proper "dose-response curve" (that is, at higher concentrations of the inhibitor, it shuts down the activity of PfSUB1 more and more effectively), which indicates that it is likely a "specific" inhibitor, instead of a non-specific compound that randomly happens to impede activity a bit for many different types of proteins (but this will have be tested against other types of proteins to know for sure). This compound, nicknamed "GF13", is a fairly weak inhibitor: at a 200 micromolar concentration, it blocks activity of PfSUB1 50%. Strong hits will block 50% of the target's activity in the 1 to 50 micromolar range (the smaller the # of micromoles per liter that are needed to shut down activity, the more potent a compound is).
Although GF13 (shown as the green surface with a cyan outline that is bound in the cleft in the center of PfSUB1, whose surface is shown in purple above) is a weak inhibitor, it is still a novel and significant hit: it provides a foundation on which we can build, and it could help us find more potent inhibitors of this potential drug target for malaria. After we write a couple papers on the GFAM results against the tuberculosis target InhA (discussed below), which could help us get some grant funding to enable additional analyses of GFAM data, we will write a paper on these results against PfSUB1. We then hope to extend this collaboration with Professor Mike Blackman's lab on this important malaria target (if we can obtain a grant to enable that extension).
GFAM experiment leads to the discovery of new hits against a key drug target for tuberculosis.
As mentioned earlier, when a drug target from Plasmodium falciparum had structural similarity with a target from another pathogen, we docked 5.6 million compounds against the targets from both pathogens. A potential drug target for malaria called "PfENR" (for Plasmodium falciparum enoyl acyl-carrier-protein reductase) has a similar atomic structure to the well-validated drug target for treating tuberculosis called "InhA". Consequently, we included the Mycobacterium tuberculosis (Mtb) enzyme InhA in our GFAM experiments against PfENR. Few pharmaceutical companies perform antibiotic (i.e., antibacterial) research anymore, which means that it is up to scientists at universities and non-profit institutes to fill that research gap. If the scientific community cannot create new ways to defeat these superbugs, then the medical community will not have the capacity to treat these drug-resistant infections that keep occurring with increasing severity, frequency, and distribution. The situation is dire. Mtb infects 8.3 - 9 million people each year, and tuberculosis kills 1.4 million people/year. A few decades ago, multi-drug-resistant tuberculosis (MDR-TB) was not a serious problem. There are now a half million new cases of MDR-TB per year. Extensively drug-resistant TB (XDR-TB) has now been found in over 92 different countries, including the U.S.A. And "nosocomial" (i.e., hospital-acquired) XDR-TB is now a growing problem. Totally-drug-resistant TB (TDR-TB) has appeared in several countries and will continue to spread. No drugs exist to treat TDR-TB.
Professor Peter J. Tonge is the first person we encountered who was willing and able to test some of the predictions from GFAM, even though it is an unfunded project. Professor Tonge is the Director of Infectious Disease Research at the Institute for Chemical Biology and Drug Discovery at Stony Brook University in NY. He has done some pioneering research against Mtb InhA and is a leading expert in the battle against XDR-TB. He offered to experimentally assess the potency of the candidate compounds we discovered in the docking calculations on GFAM that we performed against InhA. InhA (which is also called FabI) is part of a unique metabolic pathway (that is, it's an enzyme that is part of a metabolic pathway that is not present in humans), which should hopefully decrease the toxic side effects of InhA inhibitors. Specifically, ENR/InhA is part of a Fatty Acid Synthesis pathway (called "FAS II") that human cells do not have. One of the main drugs used to treat tuberculosis is called isoniazid (or "INH"), and it kills that deadly bacteria by shutting down the activity of InhA (and perhaps by also shutting down the activity of other target proteins in Mtb, as well). But drug-resistant mutants against which isoniazid loses its effectiveness keep evolving and spreading, which is why we are searching for new inhibitors of InhA that are able to counteract the main mechanism of drug resistance that allows Mtb to evade treatment with INH. Some inhibitors of InhA also block the activity of the PfENR target from malaria. By advancing the research against InhA, we might be able to simultaneously help advance the research against both totally drug-resistant tuberculosis and against multi-drug-resistant malaria.
In the results of GFAM Experiment 5, which screened the National Cancer Institute's library of compounds (that we can order for free from the NCI's Developmental Therapeutics Program), we identified 19 candidate compounds as potential inhibitors of Mtb InhA. These 19 NCI compounds were then experimentally tested in "wet lab" experiments (in test tubes and Petri dishes) by Weixuan Yu in Professor Peter Tonge's lab. Of the 16 soluble compounds, 8 candidates (at a 100 micromolar concentration) shut down InhA activity by ~ 30% or more. The most potent inhibitor we discovered displayed an IC50 value of ~ 40 micromolar (which means that when the compound is present at a 40 micromolar concentration, it inhibits InhA activity by 50%). Additional kinetic experiments were then performed on the best hits from this experiment, and the two most potent inhibitors displayed Ki values of 54 and 58 micromolar.
The binding mode predicted by AutoDock Vina for the most potent hit we discovered in GFAM experiment 5 is shown as a magenta surface, while the InhA target is shown in cyan as a ribbon above. The NAD cofactor of InhA is shown as cyan spheres. Finding new, low micromolar inhibitors of InhA is a significant achievement (and so is having a hit rate of 8/19 candidates from a virtual screen), but we will still need to optimize these compounds and make them at least a thousand times more potent before they become a drug-like candidate called a "lead". These virtual screens on GFAM (and, thus, the new inhibitors we discovered) were designed to target one of the main mechanisms that Mycobacterium tuberculosis has evolved in order to resist the effects of drug treatment with isoniazid. We are now about half-way through the process of writing a paper on these exciting new results. I hope to finish the first draft within the next month or so, and then I'll send it around to all of the co-authors to get their input and suggested revisions. After that manuscript completes the normal peer-review process, we will share the published version with all of you.
Discovery of additional new hits against a target for drug-resistant tuberculosis
The third GFAM experiment we analyzed led to the discovery of additional new hits against InhA, a key drug target for treating Mycobacterium tuberculosis, and could help seed the discovery of new drugs against extensively drug-resistant and totally drug-resistant tuberculosis.
In November of 2013, I joined Associate Professor Joel S. Freundlich's chemical biology and medicinal chemistry lab at the Rutgers University New Jersey Medical School, in Newark, NJ. Instead of having to search for collaborators in other labs to experimentally test my predictions, and then waiting many months to years to find out whether the compounds actually work against the drug targets and then against the full pathogenic organism, I can now just ask my co-workers in the Freundlich lab to help test the computational predictions. The enzyme activity assays, anti-bacterial activity assays against whole-cell Mtb cultures, metabolomics studies to investigate the different targets that are affected, synthesis of new derivatives, and medicinal chemistry-guided optimization research are all performed within the Freundlich lab. I provide the computational chemistry core that can help analyze and guide some of these efforts. I've even started learning how to do some of these "wet lab" aspects myself, to expedite the assessment of my computational predictions. In the Freundlich lab, some of our projects against Mtb involve analyzing and extending the subset of GFAM data that involve targets that can advance the treatment of XDR-TB and TDR-TB.
In GFAM experiment 9, which involved docking the Asinex library of over 500,000 compounds against PfENR and Mtb InhA, we discovered additional new inhibitors of InhA that have very novel structures, as compared to the known InhA inhibitors. In collaboration with Dr. Sean Ekins of Collaborations in Chemistry (Fuquay-Varina, NC), this project also involves the creation, testing, and development of new types of computational workflows that can filter and prioritize the docking results, in order to increase efficiency and enhance the probability of finding active, potent, non-toxic compounds. We used an innovative new combination of docking-based filters, other types of filters (which we cannot discuss until after these results are published), and visual inspection to narrow down the docking results for those 500,000 compounds and identify 20 promising candidate compounds. These 20 Asinex compounds were then tested in enzyme inhibition assays against InhA by Xin Wang, a graduate student who recently joined the Freundlich lab. Since Professor Freundlich was a medicinal chemist in the pharmaceutical industry for over a decade, he applied a more stringent bar to define what we would consider as hits: a "hit" had to inhibit InhA activity by at least 50% when the compound was present at a 50 micromolar concentration. Using this more stringent criteria, 5 of the 20 candidates are novel hits against InhA. The most potent new InhA inhibitor we discovered has an IC50 of 19 micromolar (at a concentration of 19 micromoles per liter, it inhibits InhA activity 50%).
This new hit's predicted binding mode against InhA is shown as a dark green surface, with the InhA target shown in light purple as a ribbon above. Having a 5/20, or 25%, hit rate against an enzyme from the results of a virtual screen was a great success, as was discovering another novel InhA inhibitor that was more than twice as potent as the one we found in our previous GFAM experiment against InhA. The 5 hits we discovered against InhA were then further tested for their activity against whole-cell Mycobacterium tuberculosis cultures by Dr. Mi-Sun Koo in the Freundlich lab. These "MIC99" assays (for the Microbial Inhibition Constant, or the amount of compound needed to inhibit the growth of 99% of the bacterial cells) take over a week to perform against Mtb, since Mtb grows very slowly. Because Mtb is an air-borne bacteria, these MIC assays have to be done in special "BSL3" labs (the second highest form of Biosafety Level), using the full-body "bunny suits" and respirators. Since Dr. Koo went on vacation before the assays were done, the final measurement was taken by Dr. Pradeep Kumar, of Professor David Alland's lab at the Rutgers University-NJ Medical School. Of the 5 InhA inhibitors we discovered, one of these compounds displayed potency against whole-cell cultures of the bacteria that cause tuberculosis, with an MIC99 of ~ 2.4 micrograms per milliliter.
Its predicted binding mode is shown as a light green surface, with the InhA target displayed in light purple as a ribbon. Traditional, experimental high-throughput screens against whole-cell Mtb tend to display hit rates less than 1%, but 1/5 (i.e., 20%) of the compounds we tested were active against whole-cell Mtb cultures. Professor Joel Freundlich, Dr. Shao-Gang Li, Dr. Steve Paget, and I have already designed new derivatives of the top InhA inhibitor and the top MIC hit, to test their Structure-Activity-Relationships (to figure out how changing and adding different chemical groups to these molecules increases or decreases their activity) and to try to increase their potency. These derivatives are currently being synthesized by Shao-Gang and Steve, as part of the "hit-to-lead development" process mentioned previously. After these derivatives have been made, tested in InhA activity assays, and tested in Mtb growth assays, we will write a paper on these exciting new results. Professor Freundlich, Dr. Sean Ekins, and I have already written and submitted two different grants that discuss our new approach and propose applying it to the other 5 million compounds that were docked against InhA as part of the GFAM project, and we plan to write a couple more grants that involve the Mtb subset of GFAM data. We'll let you know how it goes.
Personnel update and a request for action
Last year, the "sequester cuts" to the NIH eliminated the funding for my position in Professor Olson's lab at TSRI. Don't be sad for me - please contact your members of Congress and ask them to restore funding to the NIH! The inflation in scientific equipment and reagents rises faster than normal inflation, but the NIH budget has still not yet been restored to even the pre-sequester level. It's not even close to the level of funding it should have, considering scientific inflation, the growing problems that the medical community faces, and the fact that pharmaceutical companies keep cutting their research budgets year after year (which puts the academic scientific community under pressure to fill this research gap, or else progress against diseases will not occur rapidly enough). In addition, every single dollar from the NIH and NSF generates approximately $2 in economic output, and NIH funding has led to the creation of nearly 430,000 quality jobs. Luckily, I was able to find a great new position in Professor Joel Freundlich's lab at the Rutgers University-New Jersey Medical School, where I get to continue my fight against drug-resistant infectious diseases. But I know many scientists who are still not able to find a job.
Thank you all very much for donating your computer power to World Community Grid!!! We would not have been able to generate this mountain of data for malaria and tuberculosis research without your generous help. The massive number of virtual experiments you performed provides us with enough data to continue our research for years to come. Please be patient, and keep crunching! Many other projects still need your help.
Dr. Alex L. Perryman