Barry Bunin – Drug Repositioning Panel Talk Transcript
Okay, so I’m going to talk about collaborative platform to drive drug repurposing, which is also a platform in general for drug discovery whether repurposing or not, and one example of a collaboration that we’ve been working with the Bill & Melinda Gates Foundation has been supporting tuberculosis research and recently got nominated by researchers at the NIH’s Intramural Program, NIAID and TB Alliance, for this BioIT Best Practices award and this is interesting an example where we’re supporting tuberculosis researchers who work together worldwide and now it’s moved into a second phase with some leading TB researchers with three different big pharmas to advance those TB projects, so that’ll be one of the three case studies that I talk about.
So just briefly to get across the concept of what the main technology does, it’s called the CDD Vault® to emphasize the privacy and security for IP sensitive data prior to publication or patents. It’s definitely important to emphasize the security first, because of the collaborative capability represented by the handshake so that you can selectively share chemical structures with or without bio data or bio data without chemical structures down to an individual experiment (i.e. one molecule or IC50 measurement). So for something like drug repurposing where you want different brains and different hypothesis and predictions or even experimental validations done by different groups, this is a way to have people work as just one organization even if they’re different companies. So the broad vision of CDD is that anybody could work with anybody else with secure IT regardless of what group they’re working in, and we’ve been doing this for seven years in the cloud.. The mm4tb collaboration is an interesting example with two big pharmas, AstraZeneca and Sanofi-Aventis, working with approximately 30 different organizations and working as one cohesive group using collaborative projects in Vaults (in some cases with the big pharmas masking structures). GlaxoSmithKline shared Malaria data on CDD and Novartis shared TB data on CDD (both used the secure vault, collaborative, and public capabilities). A number of academics, for example, screening centers where they may be there for a different professors with different cloned proteins for screens with a chemical vendor library. So the screening center can use the CDD Vault without uploading the chemical library 20 different times, they just upload it once to the vault and each biologist has a separate, secure collaborative project in the vault. If you have permission to see all 20 projects, you can mine across all the data, conversely if someone else has access to just 1 project, they’d only see access that data and be none-the-wiser about the other 19 projects existance. And we’ve developed a simple way to do that, so researchers can collaborate within their natural scientific workflow.
And the twist on traditional technologies is that collaborative capability and so the point of this chart is to think of it both from the individual’s perspective as well as the organization’s needs and having temporal access, for example, if you want to double check and confirm a result before collaborating or if you want spatial partitioning like you have to see the results from three different labs, but they should only see their one-third of it and then as you heard earlier by type whether the object is a Ki or a one batch of a molecule. So that’s the place where we’ve been innovating.
And so how is this relevant to organizations that may have assets that they want repurpose with other academics or other companies with complementary capabilities? So if you think about the economics of the drug discovery process, a big company’s not going to be able to take every drug all the way forward and there are lots we heard yesterday about compounds that have a clean safety or tox profile after phase two but maybe weren’t quite efficacious enough or didn’t have a big enough market opportunity to warrant a more expensive phase three trial.
So how can you allow all that great work and potential activity maybe for a rare or neglected or even a smaller commercial indication to be taken further by another partner, and how can you figure out both from the data side, which represents intellectual property, how the groups can work together as one?
So there’s a review. This is just to give one example of the public space of a poster child for drug repositioning or repurposing for Thalidomidewhere there is data from ten other groups that we’re willing to share in the public space. Now for every one example in the public space, there’s going to be 20 times as much data in private or collaborative spaces. And so one of our colleagues, Sean Ekins, recently published this review on the repositioning of the approved drugs for rare and neglected diseases. I thought there were two interesting tables worth just sharing for example AK1, a receptor antagonist that’s been looked at for drug-resistant HIV infection and antiarrhythmic for Chagas Disease and so these were all discovered by low throughput screening methods and just give folks a chance to look at this because based on your background and knowledge of the drugs in the areas you may have other ideas.
A second nice table just is the examples discovered by higher throughput or even in Siliico methods and in particular I want to mention this Johns Hopkins Group we’re working with, David Sullivan, who does actually have a library of physical compounds that structures of course in CDD, but the actual microtitor plates available for the cost of plating and shipping he published in a Nature paper. Brian Roth also published a paper earlier on the new uses of old drugs, so this is an area which has been sort of gaining momentum for a number of years and as with the previous slides, rather than talk about each example, I’ll just give folks a second or two to think about this and be able to give them ideas for applying their own drug assets for other areas. And if folks want more information, they should just email me or Sean who’s the actual author on the paper.
So an example in that publication using CDD for drug repositioning, there’s to start with some data such as the TB screening data set using the 2D similarity search. You can substructure and Tanimoto similarity in CDD and he can do other more sophisticated pharmacophore methods on your own. You can look for matches between data sets that you have and data sets in the public sector such as set of FDA drugs and then find a molecule like your hit that is a known drug and so a project, any project which is not initially a drug repositioning, repurposing effort potentially could be!
And this is because in the public space (even though the majority of data is in private vaults and collaborative vaults) we have from the FDA site substructure searchable in CDD the information that they have on the drugs for repurposing/repositioning, as well as examples like from compound vendors or academic vendors where we have, for example, GPCR gene-family wide SAR data or lipid data from North Carolina and Colorado, so the searcher can just think outside your regular perspective that you are working on. And the reason why this is unique, is now that we’ve been doing this for seven years, we have a lot of traction with the different parts of the ecosystem, and we have this Switzerland position “if you will” where we claim no IP, our entire trust is based on the fact that we’re working with others and they get to drive it. We’ll provide the technology. Our business model is a subscription for CDD for your private use and then the public space is available for free.
So just to share two other case studies as a result of this technology and complementary groups working together and selectively sharing data. So riding the coattails of observation that Verapamil would reverse resistance for cancer therapy, malaria researchers also found that this was true. And one of the problems with malaria we heard yesterday from the presentation from SLU is that there’s resistance to Chloroquine … So if you can give a compound like Verapamil in combination, you can reverse the resistance and Professor Kelly Chibale at University of Cape Town found that this substructure of the secondary amine four atoms away from an aromatic ring would also reverse the resistance. And because he’d done a postdoc at UCSF and Professor McKerrow was working on other neglected infectious diseases, they were willing to work together and we found 79 matches with the substructure, FedExed them over and Peter Smith who has the assay in human red blood cells found a compound that would reverse the resistance and that’s important because, if you know synthesis, that can save months or quarters or years off the timeline. But the other thing that was interesting from Chris Lipinski with our first data set of the known drugs and their therapeutic indications, we found 18 matches of known drugs with this substructure and MS Discovery FedExed them over and lo and behold there were two that would almost completely reverse the resistance when used in combination – over seven-fold reversal of resistance. So potentially that could take years or decades off of the timeline. And if you were a kid with the resistant form of malaria that would potentially be quite interesting. So I share this because a) it wouldn’t happen without technology for selectively sharing data, but b) it also wouldn’t happen without connecting different people with different resources that others just don’t have available.
And so taking that to the modern scenario, to today, this is the example I alluded to earlier with Astra-Zeneca and Sanofi-Aventis with Stewart Cole who is the PI of this project and the interesting thing is that this group is working together in one Vault with selective sharing using different projects and collaborating together and so it’s sort of where I see the field going in the future.
So switching gears a little bit, there’s a lot of information to help with drug discovery that’s not that proprietary that’s even been published such as the alerts such as the Abbott NMR alerts and Pfizer Lint alerts that pick out things like aromatic nitros or sulfur compounds that you just might want to be aware of and recently we’re very proud, we co-published with Pfizer that open source models and descriptors were equivalent to the expensive commercial models, and this opens up the potential for the collective synergies between pharmas or between pharmas and start-ups and academics where you may want to profile compounds say for hERG or for human liver microsome stability without sharing the proprietary structures, just like when you anonymize patients, you don’t want any way to reverse engineer it. So we have two ways of anonymizing the information. First is the model based on the structure, so you don’t even have to share the structures, you just share the model. This poster just shows that statistically the data was equivalent to top commercial descriptors and models. This data set is particularly interesting because when Pfizer acquired Parke Davis, Pharmacy Upjohn, et cetera, they have the largest set, at least to my knowledge, of ADME and Tox data sets, so these predictions are as good or as bad as the data set that they are based on and these data sets are one of the strongest if not the strongest in the world.
And so the second part of getting commercial organizations completely comfortable is the selective sharing capabilities, so this could be used with one company, just with one other company where they may want to in-license the drug repositioning candidate or other drug candidate, but if you profile the molecule and it looks promising, that’s where you likely would want to do the next assay in the drug discovery process.
So just to summarize, people think about their individual efforts, the technology can change the game and software actually does impact the efficiency, and what we’ve done is made it so the people can collaborate if they want to without having to worry necessarily about waiting six or 12 months through legal or business development cycles and rather when something’s found (via simultaneous collaboration), you can worry more about that – and so that’s where I’d like to wrap it up.