OpenSourceMalaria:Story so far
This is a human-readable summary of the first open source drug discovery for malaria project. A less readable collection of all the relevant data can be found here.
Note: This page is currently (May 2 2012) being built, so some links are missing.
The story so far in the open source drug discovery for malaria project
Last year my lab received seed funding for a pilot project in open source drug discovery from the Medicines for Malaria Venture (MMV). Our project champion at the outset was Tim Wells, who had clearly been thinking along similar lines to me - rather than debate the idea of open source drug discovery any more, let's just try it. Jeremy Burrows quickly came on board and led the suggestion to go after a few of the actives that had been placed in the public domain in 2010 by GlaxoSmithKline and others. We got started in the lab in August 2011. We were successful in securing further funding from an ARC Linkage grant - a scheme where funds from an external agency are matched by the Australian Government. This has given us funding from May 2012 for three years - so we have enough money to drive this project for the moment to see if it works.
Naturally we're all really excited about this. The scientific idea behind the project is familiar medicinal chemistry territory - we need to find a small molecule that is effective for the treatment of malaria, and we will do that by making molecules (my lab's primary responsibility) and evaluating them (with collaborators). Based on those results, we make analogs, or ditch the series and pick another. We started with the arylpyrrole series that was one of the most attractive sets in the original GSK dataset, but there are plenty of other series that are also very attractive from a medchem perspective.
The difference with this project though (as we previously described in the 6 Laws) is that everything is open, meaning all the experiments go on the web (including the ones that did not turn out well). All the data are available. Anyone can do anything they wish with the compounds, with the proviso we are cited - the licence for the project is CC-BY-3.0, though this is sometimes not yet clear on all the various websites we use. The main difference is that anyone can take part - that people may make molecules, offer guidance and input in other ways that change the direction of the project as it is happening. i.e. rather than releasing all our data at the end of the project we release as the project is happening so that people can really become involved in the research. Thus the iterative cycle of analog synthesis in response to biological data that is normally guided by a kind of medchem intuition is now guided by the intuition of the collective. Similarly, since the biological data are all open too, it should be easier to form an objective assessment of a molecule's performance divorced from the judgement of those closest to the compounds. In the same way that in software development "with enough eyeballs all bugs are shallow" we hope that the open nature of the research makes the science better and faster. As it did with our previous synthetic project with praziquantel.
Two Rounds of Synthesis and Evaluation
Paul Ylioja started by resynthesising the two known active compounds from the GSK set (OSM-S-5 and OSM-S-6), plus a few simple derivatives, and confirming that they were active. The current list of all the compounds made thus far in this part of the project is kept in this spreadsheet. The biological evaluation was carried out by three separate labs to ensure we were on a solid footing. The original compounds contained an ester which was thought likely to hydrolyze in vivo, so various versions of the "lower half" of these leads were also evaluated to check whether the original hits were prodrugs, but all these compounds were found to be inactive. Our MMV project champion Paul Willis, who has been working closely with us ever since, recommended a few "near neighbor" compounds that also looked interesting, and we made a number of these too, which were evaluated in this first round, and one, OSM-S-9, was found to be more active than the original compounds. Sanjay Batra came on board the project and his student Soumya made some analogs varying in the position of the fluorine, though the activity of those tested to date is low. (Sanjay works at the CDRI in Lucknow, India, where Saman Habib also works - Saman is leading the Indian OSDDm project which will shortly get started). The outcome was that the original hits remained interesting (because of their reasonable potency and logP values) but that we were clearly also generating highly potent novel antimalarials in this class. Thus a second round of compounds were synthesized and evaluated, and this gave rise to several new highly potent compounds, one of which (OSM-S-39) displayed a picomolar IC50 value. This is quite impressive given the small number of compounds made to date, and is perhaps testament to the quality of the hits contained in the original GSK set.
At this point the decision was taken to take the most promising compounds on to advanced biological evaluation, to see what the promise of this class really is (rather than continue to increase potency through analog synthesis). To date the evaluations have involved:
- Metabolic assays on the two original GSK compounds and six other compounds made in this project were performed by Sue Charman's lab at Monash. The raw data are here and can be discussed here. The originals displayed good solubility but moderate degradation rates. The other compounds were degraded more slowly but at a cost of low solubility.
- One of the original GSK compounds (OSM-S-5) plus one of the most potent novel compounds identified to date (OSM-S-35) were subjected to the hERG assay and passed, perhaps implying that this class of compounds should not exhibit undesirable cardiac side effects. Discussion page here.
- Four of the compounds have also been evaluated in a late stage gametocyte assay with very interesting results indicating unusually high activity in blocking the transmission of the parasite. The original GSK compound OSM-S-5 was inactive. Discussion page here.
- However, the two original GSK compounds as well as one of the most promising novel compounds have been evaluated in mice and found to possess zero oral efficacy (spreadsheet coming as soon as cleared by creator).
What Are the Compounds Doing?
What might this series of compounds be doing to the parasite to kill it so effectively? We're not sure yet. The original screens were whole-cell assays, so while we know the compounds are effective, we don't know what they're doing in any detail. Iain Wallace from ChEMBL has done a very neat prediction of the biological role of these compounds (as well as predictions for the whole "Malaria Box", which is a set of compounds MMV are providing to people for antimalarial screening and which are the focus of a current round of Gates requests for proposals). Iain clustered the compounds as a similarity map, which is a neat way of visualizing the correlation between structure and predicted activity. Discussed also here. One of the predictions was that the compounds hit an enzyme known as DHODH, and GSK are at the time of writing screening some of the compounds against this enzyme. These predictions are made using informatics - a comparison of the structures of our compounds with other known compounds that have known activities. It's an argument based on extrapolation. We're seeking to examine whether the prediction is correct by sending a subset of compounds to Corey Nislow at the University of Toronto for examination in a yeast-based assay he's developed which does not identify for sure what the compounds are doing (which is very difficult) but provides harder biological evidence for a role.
How do we Obtain Other Compounds?
The original compounds from the GSK assay were commercially available, arising from libraries that are provided to larger companies by smaller specialised companies. In a medicinal chemistry project like this one starts with a set of compounds, then one sources further compounds that are similar. Novel compounds need to be made. Other compounds may be available by other means, however. Typically in academia we just make all the compounds in-house because we worked as a closed system, which is inefficient if relevant compounds exist elsewhere on the Earth and can be sourced by other means more quickly. In this project organic synthesis of novel compounds is currently being performed in Sydney and Lucknow. But for known compounds we need to do the following:
1) Identify Commercial Compounds: What if some compounds we require are already commercially available? How can these be found? Iain Wallace was able to do a search of databases such as eMolecules for relevant compounds above a certain threshold of similarity and filter compounds by supplier, generating the "hitlist" quickly and with no manual human input. These can be converted into spreadhseets for quick visualisation - see here for examples on Jimmy Cronshaw's series (see below for these).
2) Get Commercial Compounds: With the compounds identified, the relevant suppliers need to be contacted to ask for donations. That will probably not be trivial since it needs a human interaction. Failing that the compounds can just be bought.
3) Identify Other/academic Compounds: What about compounds that could be useful to the project but which are not commercially available, e.g. compounds sitting in academic lab fridges. Some can be found using resources like SciFinder, though these require expensive subscriptions. Many compounds that might be perfect for the project may not even be in the published literature, which is another argument for openness in science.
4) Get Other/academic Compounds: This will be a case of manual contact with interested groups. One such enquiry has already been submitted, to see how it goes. Naturally contributions are rewarded by possible authorship on resulting papers.
Ultimately, as a species, we are not very efficient at sharing valuable chemical resources. A user-driven list would be helpful - "I need compound X. I can buy it, or you can give it to me, or you can make it for me or with me, but I need the compound in timeframe Y." This is like a Molecular Craigslist, and would reduce some of the supply-demand barriers.
The arypyrrole series is the first to be examined. A forked series, the pyrazoles, looks attractive but has not yet received a great deal of input. Two other interesting-looking starting points from the GSK set, based on a triazolourea and a thienopyrimidine, are currently being resynthesised by James Cronshaw in Sydney to confirm their activity, upon which time it will be necessary to decide which, if either, are to be taken on further. In general a strategy for the project is to decide as early as possible which series are looking attractive biologically, because there are still a lot of leads to pursue. If any of these structures look synthetically attractive to you based on your previous experience, please consider joining the project.
Open Source Drug Discovery More Generally
A one-day meeting on open source drug discovery for malaria was held in February 2012. General issues surrounding the feasibility of open source drug discovery were discussed, followed by more specific malaria-related ideas. These talks are gradually going up on YouTube with annotations, and they frame many of the relevant issues, for example the landscape of drug discovery in neglected diseases, and whether patents are necessary in drug discovery. An important message is that open source drug discovery is where anyone may participate in driving the research, which is different from a more general use of the word "open" where data are made freely available, but perhaps after a delay which essentially prevents participation by others.
How We Run the Project
The way the project is run is one of the novelties, though as with everything in this project nothing is static and advice is always welcome on improvements. Raw experimental data are recorded in an online, openly-readable electronic lab notebook. The Synaptic Leap is being used to discuss ideas and results, as well as plan future work. The project's Google+ page is a light way to keep up with developments and discuss. The project's Twitter feed is a broadcast mechanism for updates. LinkedIn as used in the past on another project as a way of connecting with relevant experts, but has not been used much so far in this project. A wiki (that includes this page) is used to host the current overall project status. If you wish to participate in this project, you can sign up to all these sites, and you would then be sent the Twitter/G+ passwords so you can used the same accounts. A Facebook page is needed next but doesn't exist yet.
Why Take Part?
What of motivations? Why would people contribute? Partly to solve a problem. Partly to be involved with quality science that is open, and hence subject to the most brutal form of ongoing peer-review. Partly for academic credentials (since we'll soon be publishing a paper). Partly to demonstrate competence. Perhaps a mixture of all these things.
We're playing with the idea of a competition, though - or rather a prize. While we have certainly led the way to a very promising lead compound, we are acutely aware that there is a long road towards having a compound look sufficiently promising that it moves towards clinical trials. There's a lot of tweaking, and perhaps even the move to another series. Who knows. It's likely we will need a lot more input that we have currently been getting, and so we're playing with whether to launch a prize to stimulate input. It would be a teamless prize, awarded based on performance of individuals within a group where everything is shared. Difficult to judge, difficult to award, and hence worth doing.
A final point - the project is open. We don't own it. It exists in itself and those people most active in the project lead it. If you wish to contribute, in any capacity, please do so. There is no need to "clear" anything with me by email first. It's often the case that I will receive questions/suggestions by email. In the development of Linux, the need for Linus to approve everything caused problems, and the observation that "Linus doesn't scale". Well I don't scale either, so it's more efficient if all our discussions are held publicly. I know that a lot of people don't like this. In science we don't tend to get the idea of "beta testing" something. When data are released in science there is an expectation that the data are correct, and usually accompanied by an explanation. I don't understand this view. I'm comfortable with release of data immediately, and then I'm happy to apply a caution filter that makes me skeptical of things until repeated, or makes me unsurprised when a repeat fails.