1

White Paper

Collaborative Mobile Apps Using Social Media

and Appifying Data For Drug Discovery

Sean Ekins1,2, Alex M. Clark3 and Antony J. Williams4

1 Collaborations in Chemistry, 5616 Hilltop Needmore Road, Fuquay Varina, NC 27526, U.S.A.

2 To whom correspondence should be addressed ekinssean@yahoo.com

3 Molecular Materials Informatics, 1900 St. Jacques #302, Montreal, Quebec, Canada H3J 2S1.

4 Royal Society of Chemistry, 904 Tamaras Circle, Wake Forest, NC-27587, U.S.A.

2

Executive Summary

We are at a watershed moment for drug discovery. Can we leverage social media, collaborations and the masses of data that are in the public and private domains to accelerate drug discovery? One possible way to do this may involve methods for “appification” of data, whether in self-contained apps or those that push data to other relevant apps that can enable visualization and mining. Mobile apps that can pull in and integrate public content from many sources relating to molecules and data are also being developed. Apps for drug discovery are already evolving rapidly and are able to communicate with each other to create workflows, as well as perform more complex processes, enabling informatics aspects of drug discovery (i.e. accessing data, modeling and visualization) to be done anywhere by potentially anyone.

Analysis

The winds of change are blowing through the pharmaceutical industry creating a new ecosystem, with pharmas becoming smaller nodes in a complex network in which collaborations (with academics, CROs, public-private partnerships and not for profits) are an important component of the business model [1]. Yet, still there is an urgent need to: 1. fundamentally revamp how drugs are developed, 2. determine methods by which they can be brought to market faster and 3. provide incentives that can facilitate treatments for neglected and rare diseases. For these diseases specifically funding comes primarily from public sources, data is more open, and potential profits are thought to be non-existent. In both neglected and rare diseases, the partners are more likely to share IP because the monetary value of the IP ceases to be a barrier. So what can we do that will address these needs?

Technology development is moving faster but R&D organizations do not appear to be keeping pace as they are still wedded to the desktop computer and internet of the late 1990-2000s. The crowd is unwittingly providing us with valuable data (which we are not capturing and saving) that can be readily extracted from the web and social networks. This can enable drug safety analysis, drug repurposing and marketing by sentiment analysis using social media stream mining tools and real-time data from social networks [2] (such as Teranode, Ceiba and Swarmology). However, the availability of such tools and platforms to collect, analyze and deliver this knowledge is in its infancy, with many of them disconnected as separate islands lacking integration. This, in many ways, is analogous to what we are seeing with how mobile apps are being created and used for science as individual components with little integration [3].

Mobile Apps for Drug Discovery

The user community is demanding a new breed of chemical information software that keeps pace with the rapidly changing dynamics within the chemical industry (including pharmaceuticals). Software for drug discovery scientists has to be affordable enough for

3

all to participate, have a sufficiently intuitive user interface that becoming an expert is not mandatory, and be available anywhere, anytime.

The pace at which mobile apps have claimed a prominent position within the workflow of so many professionals is impressive. Already the capabilities of mobile devices to access, search, manipulate and exchange chemistry-related data relevant to drug discovery almost parallel those capabilities which were available on desktop computers just a few years ago. We are confident that this budding ecosystem of chemistry apps (Fig 1) will continue to grow rapidly, and that the ability of these apps to complement each other, as well as workstation-based and server-based software, will secure their place within chemical data workflows.

The modular nature of first generation mobile apps means that it is often necessary to use more than one app to accomplish a particular workflow segment, e.g. using a database searching app to locate data, and another to organize it into a collection. Passing data back and forth between apps is therefore an integral and frequent activity.

Second generation cheminformatics apps will have the facility to perform many more sophisticated functions, and in order to make ever more powerful functionality practical, these apps will need to incorporate data sharing and collaboration features as an integral part of their design e.g. QSAR data preparation and prediction, pharmacophore analysis, docking clients, 2D depiction tools for 3D data, to name but a few. Numerous additional data sharing scenarios are possible, e.g. deeper integration with online chemical databases, direct integration with electronic lab notebooks and interfacing with laboratory instrumentation via wireless communication methods. The combination of a user interface designed and optimized for the mobile form factor, cloud-based server functionality for data warehousing and extra computational capacity, and collaboration features for integration into an overall workflow, makes these projects not only technologically feasible, but in many ways preferable to traditional software.

Fig 1. Examples of mobile apps for drug

discovery

4

Mobile apps are currently much less well suited to managing big data collections than analogous desktop software, due in large part to their limited computational and storage resources, but this will change in future. Currently apps function as components: frequent data sharing is therefore a necessary part of any workflow, which is effective for small collections, i.e. hundreds of rows of data, rather than thousands or millions. Simple workflows involving big data collections, e.g. submitting a structure search to a server and fetching the best few results, are already well established. Active participation in visualization and maintenance of large data collections will require new methods for task subdivision and integration of apps within pipeline-based workflows [4].

The increased availability of data and algorithms in the cloud, accessible via standard programming interfaces, enables the first generation of scientific apps to access capabilities that require more powerful processing power. In summary, perhaps the most crucial feature for making mobile devices a viable component of a drug discovery workflow is the ability to collaboratively share molecules and data. A second generation of mobile apps is already emerging, which takes advantage of the many different technologies provided by mobile platforms that allow data to be passed back and forth between heterogeneous environments. This is potentially transformational.

Finding Apps

Apps for science and drug discovery continue to expand in number, diversity and capabilities. They may be categorized into scientific disciplines and further sub-categorized based on applications within a branch of science. As a service to the community a wiki site called www.scimobileapps.com (Fig 2.) has been set up hosting a growing list of scientific apps for all available mobile platforms. This is a valuable resource which will continue to expand in content and may be useful for the creation of future science-focused app stores.

Fig 2. An example of a mobile app description on www.scimobileapps.com.

5

Appifying data

A myriad of data and multitude of datasets for drug discovery are already available to us online but the challenge is to get them into a format that is useful. For example structure activity tables in papers and supplemental data are rarely thought of as useful outside of the context of a single publication. What if this content was made available via a mobile app or the data tweeted into an app that could mine the data and structures? As one example of appifying data, we have used the ACS GCI Pharmaceutical Roundtable solvent selection guide data (a PDF with molecule names and data) as a starting point to develop the first mobile app for green chemistry called Green Solvents (Fig. 3) that is freely available for iPhone, iPod and iPad. The ACS GCI Pharmaceutical Roundtable [5] solvent selection guide rates the listed solvents against 5 categories: safety, health, environment (air), environment (water), and environment (waste) [6]. Key parameters and criteria were then chosen for each category (e.g. flammability is one of the safety criteria). The summary table assigns a score from 1 to 10 for each solvent under the respective categories, with a score of 10 being of most concern and a score of 1 suggesting few issues. This is further simplified by using color coding with scores in the range 1 to 3 shown as green, 4 to 7 as yellow and 8 to 10 as red. This allows quick comparison between various solvents. The app was built using the Objective-C programming language, the API provided by Apple for native iOS development, and the MMDSLib library for cheminformatics functionality such as structure rendering [7, 8]. The Green Solvents app uses solvent structures grouped by chemical class as the primary point of entry. These solvents are also color coded with a brown background suggesting less desirable and a green background suggesting more desirable. The user can scroll through all the solvents and click on a molecule of interest. This opens a box which lists the molecule name, CAS registry number, scores for each category with color coding as well as links out to the ChemSpider website [9], the Mobile Reagents app [10] and the Mobile Molecular DataSheet.[11]

Fig 3. Screenshots of the Green Solvents mobile App.

6

Fig 5. Screenshots of the ODDT app on the iPhone and iPad and one of the topics

Another way to make data accessible is to tweet it into an app that enables you to mine it or perform other functions. One tool we have developed, Open Drug Discovery Teams, makes use of "tweeted" molecular data (Fig 4).

Getting Collaborations to Work – Open Drug Discovery Teams

There is potentially an alternative approach that ignores the intellectual property associated with early research in an effort to make drug discovery more open, in a manner more analogous to open source software [6]. Alongside the increasing mobility of computers the shift to mobile apps presents an opportunity to impact drug discovery [12] and specifically create Open Drug Discovery Teams (ODDT) [13]. ODDT takes advantage of the pharmaceutical data appearing in social media such as Twitter which includes experimental data, molecule structures, images and other information that could be used for drug discovery collaborations. This app can be used by scientists and the public to follow a research topic by its hashtag, potentially publish data and share their ideas in the open (Fig 5).

Fig 4. An example of tweeting molecules into the ODDT app

7

Initially we used the app to harvest Twitter feeds on the hashtags for the following diseases: malaria, tuberculosis, Huntington’s disease, HIV/AIDS, and Sanfilippo syndrome, as well as the research topic green chemistry [14] which is of interest because its community is highly receptive to open collaboration. We have since added Chagas Disease, Leishmaniasis, H5N1 bird flu and Giant Axonal Neuropathy. All of these subjects have high potential for positively impacting the research environment using computational approaches and dissemination of information via mobile apps [15, 16]. We have used Twitter to feed content into these topics, by providing links to molecules and links to structure-activity tables.

We have also added the ability to endorse or reject documents by emitting a personal tweet with an encoded directive. We gather thumbnail images for each document, by parsing HTML files, and pre-analyzing molecular data such as molecular structures (2D and 3D), reactions and collections of structures and data.

More recently we have started to populate the app with documents summarized by Google Alerts,[17] and started a crowdfunding campaign using IndieGoGo [18] to assist in the integration of additional data sources.

Future versions of the software could integrate with other cheminformatics and drug discovery related apps (e.g. structure searching, activity data extraction, structure-activity series creation, automated model building, docking against known targets, pharmacophore hypothesis generation etc.).

Drug Discovery Teams

For organisations that want to merge their proprietary data with public data one could imagine using the ODDT app with a modified version of the server, designed to work with non-public data sources. We will leverage the ongoing development of software for analysis of documents and chemical data to provide informatics capabilities for content discovery and extraction.

Conclusion

The appification of drug discovery data and the potential for using social media for

collaboration lowers the barriers to participation and potentially enables anyone to

become involved in drug discovery, anywhere.

Acknowledgments

The photo for Sanfillipo Syndrome in the ODDT app is courtesy of Jill Wood www.jonahsjustbegun.org

8

References

1. Ekins, S., et al., Three Disruptive Strategies for Removing Drug Discovery Bottlenecks Submitted 2012.

2. Martens, D., B. Baesens, and T. Fawcett, Editorial survey: swarm intelligence for data mining. Mach Learn, 2011. 82: p. 1-42.

3. Cooper, S., et al., Predicting protein structures with a multiplayer online game. Nature, 2010. 466(7307): p. 756-60.

4. Clark, A.M., S. Ekins, and A.J. Williams, Redefining cheminformatics with intuitive collaborative mobile apps. submitted, 2012.

5. American Chemical Society Green Chemistry InstituteTM Pharmaceutical Roundtable [cited; Available from: www.acs.org/gcipharmaroundtable.

6. Williams, A.J., et al., Current and future challenges for the collaborative computational technologies for the life sciences, in Collaborative computational technologies for biomedical research, S. Ekins, M.A.Z. Hupcey, and A.J. Williams, Editors. 2011, Wiley and Sons: Hoboken, NJ. p. 491-517.

7. Molecular Materials Informatics. [cited; Available from: http://molmatinf.com/mmdslib.html.

8. Clark, A.M., Basic primitives for molecular diagram sketching. J Cheminform, 2010. 2(1): p. 8.

9. ChemSpider. [cited; Available from: www.chemspider.com. 10. Mobile Reagents. [cited; Available from: http://mobilereagents.com/. 11. MMDSLib. [cited; Available from: http://molmatinf.com/products.html#section14. 12. Williams, A.J., et al., Mobile apps for chemistry in the world of drug discovery. Drug Disc

Today, 2011. 16: p. 928-939. 13. Ekins, S., A.M. Clark, and A.J. Williams, Open Drug Discovery Teams: A Chemistry

Mobile App for Collaboration. Submitted, 2012. 14. Anastas, P.T. and J.C. Warner, Green Chemistry: Theory and Practice. 1998, New York:

Oxford University Press Inc. 15. Ekins, S., A.M. Clark, and A.J. Williams, Incorporating Green Chemistry Concepts into

Mobile Apps: Green Solvents. Submitted, 2012. 16. Ekins, S., A.M. Clark, and A.J. Williams. Communicating green chemistry by mobile

apps The Nexus Newsletter 2011 [cited; Available from: http://portal.acs.org/portal/fileFetch/C/CNBP_027943/pdf/CNBP_027943.pdf.

17. Google Alerts [cited; Available from: http://www.google.com/alerts. 18. Ekins, S. and A.M. Clark. Open Drug Discovery Teams; Crowdfunding. 2012 [cited;

IndieGoGo]. Available from: http://www.indiegogo.com/projects/122117?a=701254.

Further Information

Please contact us for further details or suggestions at: aclark@molmatinf.com and

ekinssean@yahoo.com

You can learn more about the ODDT app at: http://www.scimobileapps.com/index.php?title=Open_Drug_Discovery_Teams And frequent blogs at http://www.collabchem.com and http://cheminf20.org/