Video presentations of ACS, San Diego are now online

If you have missed our presentations from the 243rd ACS Nation Meeting in San Diego, you can watch them online here:

1. Mcule.com: A public web service for drug discovery
2. Registration system of mcule: InChI is the key

You can check about 400 other presentations online here.

Metformin for treating blindness

I found this nice example for polypharmacology on Science Blog about a new indication of metformin.

Here is metformin:


The story in a nutshell:

"University of Texas Medical Branch at Galveston researchers have discovered that ... metformin, which is commonly used to control blood sugar levels in type 2 diabetes, also substantially reduced the effects of uveitis, an inflammation of the tissues just below the outer surface of the eyeball. Uveitis causes 10 to 15 percent of all cases of blindness in the United States. The only treatment now available for the disorder is steroid therapy, which has serious side effects and cannot be used long-term."

I found an ancient article suggesting that metformin acts as a weak histamine agonist, but only histamine H1 and H2 receptors were available at that time. This paper shows that metformin can increase gastric acid levels which is probably associated with a weak H2 stimulation. It would be interesting to see if metformin has got a significant level of H4 affinity. Since H4 antagonism has been shown to reduce inflammation, H4 affinity of metformin might be the missing link here.

In fact, we and others have already found several guanidine containing H4 ligands, see some examples here:

This compound was found by our large-scale structure-based H4 screen.

This is agmatine, published as a low affinity H4 ligand in this paper.

This is VUF8430, another H4 ligand published in this paper.



Anyone interested in measuring the H4 affinity of metformin? :)

Software for all at mcule.com

In the previous blog post, I commented on some aspects on the “Federation of Independent Researchers” – an interesting initiative for smaller players of the pharma/biotech industry. Among the comments on the original post in Derek Lowe’s In the Pipeline, there were a few about software needs of individuals and small companies. For example:

"I'm intrigued by the idea relating to computational chemistry software and finding a way for small companies, particularly startups, to get access to sophisticated modeling and docking software."

Sophisticated modelling tools are expensive. No question. In fact, industrial, annual subscriptions range from $5-200k. Open source is free. So it is quite logical to suggest putting together a software package from open source modelling components. As the comment follows:

"… all the underlying force fields and QM models have been published … it would just take a team of dedicated programmers and computational chemists time and passion to create it"

There are several passionate open-source chemoinformaticions with great expertise, so this part is OK. But time is always an issue as pointed out by Rajarshi Guha:

"It just needs somebody with the time and expertise to implement them. And the combination of these two (in the absence of funding) is not always easy to find."

The other problem is that open source tools are generally not developed systematically enough to provide a complete solution. The development is typically governed by the contributors’ (academic) projects and open-source codes are generated on the side for problems the contributors have to solve anyway. Because of that there always will be missing components. So it looks that providing a complete solution would need several passionate developers working on this full-time, systematically. Putting together the pieces, adding some glue where needed and writing the missing components to complete the jigsaw puzzle. And this is what mcule.com is doing: we integrate. This is a large jigsaw puzzle though. In the absence of funding, this needs a business model. Many people have asked what’s the business model behind mcule.com? So here are some thoughts:

The puzzle wouldn’t be complete without the commercial tools that have been developed for years and reached a level that makes them superior for several tasks compared to open-source ones. We negotiate good prices with software developers and provide them at mcule.com on a subscription basis.

1. First main difference from standard commercial tool licenses is that we provide subscriptions for 1, 3, 6 and 12 months. This will allow people to subscribe for a tool for a single project only and don’t need to pay for an annual license. We think that this will attract smaller companies and individual consultants, who can’t afford maintaining long-term licenses, but want to get tools for single projects. This will be possible at mcule.com.
2. We provide full IT infrastructure: you do the clicks at mcule.com, we run the calculations automatically on the cloud. No hardware investments, no maintenance costs, no need paying for system administrator, etc.
3. Licenses are not CPU limited. What we limit is the maximum number of molecules that these tools can be applied to. This is much more calculable than the number of CPUs. Let’s imagine someone wants do make a large-scale docking. I’m not sure he/she will be able to calculate how long the calculations will run on X number of CPUs. But he/she will definitely know better how many molecules will be screened.

One commenter on the thread said:

"I'm thinking of some kind of virtual server, or remote desktop style operation. Your individual contractor can connect from wherever, and have full access to a range of tools, then transfer their data back to their own location for safekeeping. You would need some kind of central server farm somewhere, but this could probably be hosted on one of the increasing number of cloud services floating around the net these days."

Looks like we can read thoughts. This is exactly what we do.

This is a whole new business model though, not just for us, but most importantly for software vendors. So then comes the question from one of the commenters on Derek Lowe’s blog post:

“Can we propose an alternative business model to software vendors?”

Honestly, we weren’t sure about that at the beginning. But now we can say: Yes, we can! We are very close to sign agreements with some of the big players on the modelling software market. Why is this interesting for them? For various reasons. Most importantly, they were unable to collect long tail users so far. Big pharma has a large budget, can make long-term decisions, has the IT infrastructure in-house, people for maintenance, etc. What’s available from these on the other side? None. So what will a start-up biotech say to an offer for a single tool license alone for $20k? No, thanks. So, how about this other offer for the same tool for a single project (1 month license), complete IT infrastructure, unlimited CPU, no maintenance, no installation, ready to use for $5k? I think that’s something that can work, but we will see what our users will say. I think it is competitive with maintaining significant IT resources, spending days with finding free tools, installing them, writing the missing components, etc. and it is definitely competitive with buying an annual license alone for $20k.

I really liked the renting idea of DeepDyve by making scientific papers available for a few days - an alternative of annual journal subscriptions. I think mcule provides something similar by offering 1 month subscriptions for software. Plus we offer a lot more. Here are the plans:

1. Open community

Since we use some resources developed by the open-source community, we try to give back something. So, people can use open-source tools with some limitations free-of-charge at mcule.com taking that the resulting molecule collections will be public too. This might be an option for people working on non-profit projects e.g. on neglected diseases.

2. Academic users

Significantly reduced prices compared to industrial ones. Resulting collections can be private.

3. Small companies, consultants

Short-term licenses and different bundles/packages can be attractive for small industrial users. We also remove significant burden from their shoulders by providing the IT infrastructure and everything as ready-to-use.

4. Big pharma

Long-term licenses are also available. We provide enterprise solutions to adjust this plan to the actual needs of the companies. They might be interested in the following components: large range of tools, validated screening workflows, large purchasable compound library and easy, single package compound ordering.

What do you think about the above plans? Would be any of these plans interesting for you? I would be very interested to hear opinions!

Besides software, all the users get access to a high quality, up-to-date molecule database, which is the other main component of mcule.com. I will write about this in more detail in the next post!

ACS vs. SCA

I read an interesting blog post at Derek Lowe’s In ThePipeline about a proposal for a “Federation of Independent Scientists”.

Some background: drug discovery and thus the pharma industry are in big trouble (we all know about the increasing costs and small number of approved drugs). Taking the extremely large economical contribution of the pharma industry the consequences are very serious. Unfortunately, drug discovery projects typically span 10-15 years and therefore even if we change something at the front, we won’t see the difference for a while at the end. Nevertheless, everybody agrees that something should be done differently. Different companies give different answers to the problem. Some of them, like AstraZeneca and Pfizer, chose to shut down several sites to cut back costs, thus increasing the number of unemployed scientists and this is the point where this gets a social problem. Since the chances to find a place at another big pharma is relatively low these days, and the number of available academic positions is limited, some of these people join to small biotechs, or found a company themselves or become consultants. While these plans have several advantages, being alone or playing in a small team always limits the resources for getting the job done. Thus the proposal of Mrs. McGreevy:

“What about a voluntary association of independent research scientists?”

Proposed names for this association were: "Federation of Independent Scientist" or "SCA - Society for Chemistry in America" (opposite of ACS).

Members of the association could basically get:

1. Group rates on health and life insurance
2. Group rates on access to journals and library services
3. Online community for support and networking
4. Support for grant writing
5. Marketplace (advertising and bidding for contracts)
6. Special rates for other resources like HTS libraries

I think using the group power to negotiate with suppliers on the above products/services is definitely a good idea. On the other hand, it looks to me that ACS has already addressed most of these points, at least they are offering group rates for insurances, ACS publications, plus free networking, and career opportunities. To make it clear, I’m not saying these problems have been solved already. In fact, there are a lot of non-ACS journals and access to the ACS publications is still too expensive especially for unemployed people. All I’m saying is that before starting a new association, why not trying to urge ACS to make further steps? Building up a new association is hard work, and unemployed people don’t have years to wait for this to be evolved. ACS probably has enough power to negotiate with suppliers of any kind, and probably can provide better discounts of his own publications if it is forced. The question is: do the supporters of the idea have enough power to persuade ACS to make such changes? ACS currently has about 160,000 members. I think if this initiative can gather ~10,000 supporter members, then it can make a difference. I'm maybe a little naive, but if the number of demanders will be large enough, I don’t think ACS could ignore them. 

Besides this, several of the commenters mentioned already available solutions for some of the listed problems. In particular, DeepDyve can be an alternative for annual journal subscriptions. DeepDyve provides a renting service of scientific papers for $0.99. Renting means: viewing is allowed, downloading is not. While the list of accessible journals is impressive, I don’t see any ACS journals. Again it might sound naive, but why not urge ACS and DeepDyve to start negotiating and make an even more attractive renting service model for ACS members?

Some of the commenters brought up another major problem, namely the need of modelling software. Most of the small biotech companies and also consultants need some tools to work with, but prices of commercial software are high and it is not the only expense here: hardware infrastructure, maintenance, data and software integration, etc. Looks like mcule.com was a good idea! I will write about our solution for these problems in an upcoming post. Stay tuned!

Some thoughts about ACS in San Diego

I attended the ACS a few times before, but this one was special. Not just because this time the meeting was held in San Diego (it was great to return to this beautiful city and meet some friends and colleagues I haven’t seen for a while), but even more importantly we announced the release of the beta version of mcule.com (go to mcule.com and enter your email address to get access). We put down the basics. Now the question was where to go from here. We certainly had a lot of ideas about where we could go, but we needed (and still need) some feedback from our future users on these plans. In general, feedback was positive about the beta version and people found some of our plans attractive. This is good.

People already tested mcule.com will probably see the difference from other chemical database hosting websites, but those who haven’t seen it yet were asking why are we different. So here is why:

1. First, we provide purchasable molecules AND screening tools. There are some services for either of the two individually, but such an integrated system of these two components makes mcule.com unique.

2. Data quality: we have spent several months developing a rigorous registration system yielding in a high quality database. These days, when large chemical and biological data are accessible, emphasis is placed on the quality rather than on the quantity of databases.

3. Top IT technology behind mcule.com allows many things that are not available anywhere else. For example, flexible collection management for millions of compounds is not a trivial task at all but it is possible on mcule.com.

Another very important feedback coming from many computational experts and non-experts was the following:

- You provide high quality, purchasable molecules. OK. And you will provide a bunch of searching/screening tools. Nice. Now, how will I know which tool is the best for a particular problem?

So it looks like providing a large haystack of molecules and different pitch forks might be attractive for experts, but even they want to know what the best tools for finding the needle are. Computational chemists especially working in the pharma industry simply don’t have time to make large-scale assessments of tools. Looks like we have to do this extra step. And we are happy to do that! So therefore, one of our primary focuses for the next few months is to build up automated screening workflows, optimized and validated by using reference molecules. We have already done this for a few case studies even with experimental validation. For example you can check out our ACS poster on GPCR fragment library selection with 100% hit rate. Such validated, preset workflows will be provided for our users. We will also allow automated validation of screening workflows on any target for which reference molecules are available to assess their predictive power. The ultimate goal is something like this: user goes to mcule.com, enters a target name and gets potential ligands coming out of a validated virtual screening workflow. We are working on it …

PS: You can find all our ACS presentations here.

We are going to the ACS Meeting!

The 243rd ACS National Meeting & Exposition is one of the world's largest scientific conference of the year held in March 25-29, San Diego. It brings together academics and practitioners from all over the world and provides a venue for exchanging, learning and exploring new ideas. We are excited to be part of this great event!

We will give 2 oral presentations.
One is about mcule.com, explaining what mcule can offer to our users and suppliers, focusing on the integrated system of screening tools and cloud technology.

In the other presentation we would like to summarize how InChI is implemented into the mcule registration system and how it is used effectively with our supplier database and open registration services.

In addition, we will present 2 case studies of using drug discovery tools on mcule.com, with the help of the following posters:

1. Seeking novel JAK1 inhibitors

Janus kinase 1 (JAK1) is a human tyrosine kinase protein critical for cytokine signalling. It was recently reported that JAK1 plays a critical role in metastasis formation by enabling cell contractions. This makes JAK1 a very attractive cancer drug target. The available JAK1 crystal structures prompted us to set up a structure-based screening protocol against JAK1. The mcule supplier database and the molecular library of the National Cancer Institute (NCI) were screened against JAK1 by using this protocol. Several promising JAK1 inhibitor candidates were identified by these virtual screens and the most interesting compounds will be tested by biological in vitro assays.

The virtual screens were conducted using cheminformatic tools available at mcule.com and are accessible for everyone who is interested. Mcule also provides several other virtual screening tools and filtering techniques making the screening protocols flexible and customizable.

2. GPCR fragment design

Several class “A” G-protein-coupled receptor (GPCR) crystal structures have been recently published. Similar structural elements of these GPCRs suggest that GPCR ligands might share structural similarities. Mcule developed a virtual screening protocol to assess GPCR-likeness of candidate molecules by structure-based docking. Screening criteria of GPCR-likeness was set by using GPCR bioactivity data from the ChEMBL database. The GPCR-likeness of fragmented druglike molecules from the ChEMBL database and a GPCR library of our chemistry partner were screened by the developed protocol. Ligands with high GPCR-likeness were selected for further pharmacological characterization.

The GPCR-likeness assessment protocol was developed by using cheminformatic tools available at mcule.com and it will become accessible for the public to assess the GPCR-likeness of subsets of the mcule database or external libraries. Mcule provides several other virtual screening tools and filtering techniques making the GPCR-likeness assessment protocol flexible and customizable.

We expect many people to be interested in our projects!

See you there and stay tuned for more information on twitter and facebook!

What's going on ...

The mcule project was started in last June with very ambitious goals. Several months have passed and you could ask what’s happening behind the scenes? And when can I finally start searching/screening/ordering etc. Well, a lot of things happened since last June.

Laying down the foundations took more time than we expected. But the good news is: we are almost there now. So here is a short summary of what we have done and what is currently being done, and most importantly: what you can expect in the upcoming first releases of mcule.

Fundamental rules for the registration system have been set and implemented. This will guarantee that our vendor database will contain high quality molecules, and ID ambiguities will be close to zero. If you read the last posts of this blog you might be aware that our molecule registration system is based on InChI. While it is a very powerful tool for structure registration, it needs some adjustments to make sure no corner cases are missed. Vendor companies provide their structural data in SDF format, which suffers from far more limitations. It seems that no molecule representation will be sufficient on its own to represent the whole chemical space (or even just the accessible part of it). It is just too complex. Another source of ambiguities derives from drawing errors. SDF writing rules of molecule sketchers (e.g. setting the chiral flag automatically) can also result in potential stereochemical ambiguities. What we can do is to define specific structural checks to filter out potentially problematic molecules. By inspecting these problematic cases, new preparation steps can be defined to facilitate automatic registration. We have now more than 50 such checks/preparations in our registration system.

First large scale test of our registration system involving 2.5 M compounds has been completed. Results are promising, 97% of the molecules could be automatically processed, while only 3% of the molecules were subjected to further inspections. With some slight changes we can further reduce the number of retained molecules down to 0.5 - 0.1%. We are doing the final touches and will run the registration on the first few millions of vendor compounds again. This will hopefully provide you a reasonable compound deck to start your searches on.

At the same time, the first searching/screening filters have been integrated with the test database. Simple searches (exact, similarity, substructure) will be available in the first release, while more complex searches (e.g. docking, pharmacophore searches) will appear among the filters soon. You will be able to feed these engines with one of the leading javascript editors, alternatively you can search for MCULE ID, InChI or SMILES.

As to the web interface, we hope that you will get an experience, which will make mcule.com your default searching/screening tool. You get (i) informative molecule index pages containing (among others) molecule IDs, properties, vendor and ordering information, (ii) drag and drop filters to build more complex screening workflows, (iii) flexibly manageable molecule collections/hitlists quickly displayed in list/grid views, and more.

We expect to send out the invitation for the private beta version soon, so please stay tuned!