As one of the research managers in a large research hospital, you are in charge of neurobiology studies. You have been asked to work with the manager of the oncology studies to potentially implement a

Crowd sourcing in drug discovery Crowd sourcing is emerging as an open-innovation approach to promote collaboration and harness the complementary expertise of academic and industrial partners in the early stages of drug discovery. Here, we highlight examples of such initiatives and discuss key success factors.

Greater collaboration between academic institutions and the pharmaceutical industry is increasingly being pur - sued to access and foster innovation in the early stages of drug discovery. The hope is that such collaborations could help to address the need to improve research and development (R&D) productivity in industry 1, and also enable academic institutions to more effectively exploit the translational potential of their research. Although there are a growing number of examples of industry collaborating closely with a major academic partner, approaches that harness the expertise of a larger sci - entific community to address a specific question have been more limited. Nevertheless, using the internet as a platform, an open-innovation model known as ‘crowd sourcing’ 2 is now being tested in early-stage drug dis - covery by several organizations. This model has suc - cessfully been used in other sectors (for example, the Procter & Gamble connect and develop portal ( www. pgconnectdevelop.com ) through which consumers can put forward their ideas for product improvements or novel products).

Applications of crowd sourcing Originally, crowd sourcing was defined as a mechanism by which specific problems are communicated to an unknown group of potential solvers in the form of an open call, usually via the internet; the community (the crowd) is asked to provide solutions and the ‘winners’ are rewarded. In 2001, Eli Lilly was the first company to introduce this concept in drug discovery, with the estab - lishment of the InnoCentive platform ( www.innocen - tive.com ). Organizations in need of answers (seekers) post specific questions (challenges) on an internet mar - ketplace. The web community can then provide solu - tions to the challenge (solvers). In each challenge, the seeking company can select the ‘best’ solution and the winning solver transfers the intellectual property (IP) to the seeker in return for a financial reward. InnoCentive is now an independent organization with a solver com - munity of more than 200,000 experts from more than 20 countries. Further crowd sourcing initiatives in the area of early- stage drug discovery have followed in the past 2 years (see Supplementary information S1 (table) for more details of the examples discussed below). Importantly, in contrast to the classical concept of crowd sourcing, in which the task is finished once the solution has been provided, the goal of these initiatives is to seek novel ideas that are then pursued further in a more collabora - tive approach. The key benefits are that potential solvers, most of whom are researchers in academic institutes or small companies, gain access to specific tools or knowl - edge, such as data, assays, compounds or drug discovery expertise in large pharmaceutical companies, whereas the searching organization gains novel ideas, targets, compounds or tools (such as novel assays or models) that help to address a specific challenge. For example, in May 2009 Bayer Healthcare intro - duced its Grants4Targets initiative (G4T; www.grant - s4targets.com )3, the goal of which is to discover new therapeutic options by bringing together knowledge on potential novel targets in academia with drug devel - opment expertise within the company. Grants for the validation of innovative targets in oncology, cardiology, molecular imaging and gynaecology are provided for a period of 1 year. To promote the validation process and generate value for both partners, senior scientists from the company are appointed to support the projects with expertise and tools. Three types of grants are provided (support, focus and collaborative), and IP remains fully with the applicants for the support and the focus grants 3. After the grant period, promising targets may be pursued further via collaborative agreements. So far, four calls have been completed, more than 380 grant applications have been received, and 59 grants have been awarded to academic groups worldwide 3. A similar approach was recently initiated by MRC Technology (MRCT), the technology transfer arm of the Medical Research Council (MRC) UK. MRCT’s Centre for Therapeutics Discovery is focused on ‘de-risking’ early-stage academic projects in areas of substantial unmet medical need, showing their potential in pre - clinical models before partnering with industry. This resource was originally developed to translate MRC- based research, but in 2010 they started their Call for Targets programme ( www.callfortargets.org ) to identify early-stage projects from non-MRC sources. The review process has two stages: an initial triage, which is followed Monika Lessl and Khusru Asadullah are at Global Drug Discovery Bayer HealthCare Pharmaceuticals, Muellerstraße 178, 13342 Berlin, Germany. Justin. S. Bryans is at the Centre for Therapeutics Discovery, MRC Technology, 1–3 Burtonhole Lane, London NW7 1AD, UK. Duncan Richards is at GlaxoSmithKIine, Academic Discovery Performance Unit, Medicines Research Centre, Stevenage SG1 2NY, UK. Correspondence to M.L.  e-mail: Monika.Lessl@bayer. com COMMENT NATUR E R EVIEWS | DRUG DISCOVERY VOLUME 10 | APRIL 2011 | 241 © 2011 Macmillan Publishers Limited. All rights reserved by a full proposal that is reviewed in conjunction with an external panel of experts. Key criteria include target novelty, medical need, disease association, reagent avail - ability and structural data. So far, out of 112 proposals received, 52 projects have passed the triage and out of the 17 heard at full application stage, 13 have been accepted and collaborative projects have either started or are under negotiation. Projects are progressed as true collaborations between MRCT and the academic group and, impor - tantly, no transfer of IP rights is required — any IP that is developed as part of the collaboration is jointly owned and any revenue that is generated is split between the two parties under the terms of a pre-negotiated agreement. Aiming for novel compounds rather than targets, Eli Lilly introduced its Phenotypic Drug Discovery pro - gramme (PD 2; www.pd2.lilly.com ) in June 2009 as a platform to search for novel chemical compounds with high structural diversity. Structures can be submitted via the internet and are evaluated by specific cheminfor - matics tools to determine whether they meet Eli Lilly- defined minimum criteria, such as structural novelty or drug-like properties. If a compound passes the first-line evaluation, Eli Lilly analyses the compound in a panel of phenotypic screens to evaluate whether the compound has activities linked to its key indications (Alzheimer’s disease, cancer, diabetes or bone formation). If the screen provides promising data (which are provided freely to the investigators), the compound can be evaluated fur - ther through either in-licensing or collaboration. Another approach has been taken by GlaxoSmithKline with their Pharma in Partnership programme ( www. pharmainpartnership.gsk.com ), in which the company posts information on assets that are either potentially suitable for repurposing because they show clear phar - macology but have perhaps failed in their initial indi - cations, or that are novel targets for which additional information is required to validate the target or identify the optimal clinical indication. Academics are invited to register for updates as new molecules or mechanisms are posted, and to submit proposals to address the key question posed for that asset. The potential outcomes for successful proposals are flexible, ranging from a material transfer agreement for an in vitro study to full adoption as a GlaxoSmithKline clinical programme. One particu - lar expectation is that the successful proposals will lead to research collaborations with a substantial contribu - tion from the proposer. The programme was launched in April 2010, and since then five molecules/mechanisms have been posted. The first successful proposal has been identified and a collaboration is under negotiation.

Factors for success To use crowd sourcing successfully, it is critical that the questions or challenges to be addressed are suitable, precisely defined and clearly presented, and that what is expected from potential solvers and offered by the searching organization is clearly communicated. In the examples mentioned, the organizations are seeking novel targets or compounds in clearly defined indications, or additional therapeutic options for specific, well-defined compounds. Use of specific questions within the submis - sion template can guide the proposer to ensure that they are addressing the key questions. A precise definition of search terms (for example, ‘what is a target?’) avoids futile efforts. This also means that organizations have to be sufficiently transparent in explaining what they are interested in, and they have to overcome concerns regarding confidentiality of the information provided. Operational implementation is another key success factor. An accessible internet submission tool that pro - vides a simple and transparent submission process is needed. A good example is the flow chart of the PD 2 initiative. The transparency and speed of the evaluation procedure are important. In the case of the G4T pro - gramme, applicants receive feedback on the success of their application within 8 weeks of submission, and for Pharma in Partnership, feedback is sent within 28 days of the call closing. Low bureaucratic hurdles for both partners are also needed to make submissions attractive and manageable, and the IP policy should also be clearly defined. Pre-negotiated master contracts can facilitate rapid execution of agreements. To generate awareness of the call, the initiative has to be communicated appropriately; for example, through scientific conferences, targeted advertising and mail - ings to relevant institutions, societies and field leaders.

Finally, it is essential that proposer organizations are willing and ready to take up externally generated ideas.

Collaboration with external parties needs particu - lar attention to relationship management and usually requires dedicated resources. Crowd sourcing initiatives in drug discovery are still in their infancy, and whether they will have a substan - tial impact has yet to be proven. Nevertheless, an initial evaluation of the impact of the InnoCentive initiative was promising 4, and our own experience so far suggests that such initiatives could provide an important source of future innovation in early-stage drug discovery. Monika Lessl, Justin S. Bryans, Duncan Richards and Khusru Asadullah 1. Paul, S. M. et al. How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nature Rev. Drug. Discov. 9, 203–214 (2010). 2. Howe, J. (ed.) Crowdsourcing: Why the Power of the Crowd is Driving the Future of Business . (Crown Publishing Group, New York, 2008). 3. Lessl, M. et al. Grants4Targets — a novel approach to translate basic research to novel drugs. Drug Discov. Today 1 Dec 2010 (doi:10.1016/j.drudis.2010.11.013).4. Bishop, M. The total economic impact of InnoCentive’s enterprise solution: challenges, InnoCentive@work and ONRAMP. (Forrester Business Consulting, Cambridge, Massachusetts, 2010).

Competing financial interestsThe authors declare competing financial interests : see Web version for details.

AcknowledgementsThe authors thank S. Schoepe for her valuable contributions in editing the manuscript. 242 | APRIL 2011 | VOLUME 10 www.nature.com/reviews/drugdisc COMMENT © 2011 Macmillan Publishers Limited. All rights reserved