Global Pharmaceutical Company and AI Drug Discovery

Project Objective

Through the Columbia Business School and the Vagelos College of Physicians and Surgeons masterclass, Healthcare Management, Design and Strategy, our team was able to work closely with a Global Pharmaceutical Company, which we will refer to as GPC, to address one of their fundamental business problems: how to participate in AI enabled drug discovery.

The specific question that our team focused on was: What is GPC’s current participation model in AI drug discovery relative to a peer set of competitors and where is there room for GPC to take action? Our goal in answering this question was to provide GPC with a recommended approach for participation in AI drug discovery.

Problem Background

While drug discovery is crucial to drive impact at any pharma company, it is becoming increasingly more expensive. On average, it costs more than $2.5 billion to develop a single drug over ten years. As a result, pharma companies have increasingly looked to small biotech companies and academic centers for innovations to bolster internal R&D efforts. In the last several years, there has been a new towards integrating AI to support drug discovery. By engaging with AI drug discovery companies, pharma companies are hoping to reduce the costs associated with drug discovery and increase the speed and accuracy of finding new drugs for development.

Solution Approach

We utilized the choice structuring approach to address this fundamental business question for GPC.

Framing Choices & Brainstorming Possible Options

We structured our project around four sub-questions that would enable us to frame the choices and identify possible options:

1. How do other pharma companies participate in AI drug discovery?

2. What are trends across AI drug discovery companies targeted for participation?

3. What is the impact of participation for pharma companies?

4. Should GPC change its participation in AI drug discovery and, if so, how?

To address these sub-questions, we built a benchmarking dataset that captured publicly available AI drug discovery engagements over the last ten years. We focused this dataset on GPC and a peer set of five similar organizations. We identified over 50 AI drug discovery engagements, and the dataset enabled us to track information about pharma participation methods and objectives, the AI drug discovery target companies, and deal outcomes. Importantly, we were able to categorize pharma participation into the below models.

In analyzing our benchmarking dataset, we generated key takeaways to answer our sub-questions. In response to (1), we realized that most engagements were conducted through collaboration agreements. Other common participation methods included research collaborations, equity investments, in-house R&D hubs / teams, and incubators. In response to (2), we identified that most AI companies targeted have a unique AI drug discovery platform or model, focus on small molecules or biologics, and work across all therapeutic areas. To better understand (3), we sought to validate the hypothesis that pharma companies were engaging with AI companies because they wanted a cheaper and faster drug discovery process. However, given the lack of deal maturity in the space, we found that limited outcome data (i.e., number of drugs launched or discovered) exists, which suggests that the impact of AI drug discovery is difficult to measure based on publicly released information alone. Lastly, in addressing (4), our public research suggested that GPC is significantly less active than its peers in AI drug discovery. This insight led us to believe that there was room for GPC to add to its current participation model in AI drug discovery. Our next step in the process was to figure out how GPC could adapt its participation model by defining options and specifying conditions.

Options & Conditions

To determine the options and conditions, we led a discussion with GPC stakeholders to understand how GPC internally thinks about engaging in AI drug discovery and makes decisions about participation methods. Based on these discussions, we identified four different options that combine a unique mix of participation methods.

We utilized our initial stakeholder discussion as well as subsequent stakeholder interviews to build the conditions, which were primarily focused on resources, cost, organizational structure, integration, speed and volume of access to novel AI drug discovery technology and expertise, and GPC’s external positioning as a premier research-intensive biopharma company.

Barriers to Choice & Validation

From our stakeholder conversations, we identified the conditions that we felt least confident were true. To validate the conditions, we took a two-pronged approach. First, we utilized input and feedback from GPC stakeholders to understand whether conditions held true. This input was essential given the GPC stakeholders would be responsible for implementing GPC’s AI drug discovery participation plan. Second, we utilized case studies from our benchmarking dataset to test whether conditions held true. This case study hypothesis-testing provided a useful external perspective. Overall, this validation approach enabled us to rule out conditions that did not hold true and to make an informed choice about the recommended option for GPC to pursue.

Recommendation

Our final recommendation was for GPC to continue to participate in AI drug discovery through collaborations agreements and research collaborations, equity investments, licensing, and add a startup incubator with AI drug discovery as a focus area (i.e., Option D). Given GPC’s existing expertise, our team focused on providing GPC with a proposed structure for an incubator. Based on two case studies of benchmark pharma companies’ incubators, we identified important inputs and outputs for an incubator. We also identified methods for integrating AI drug discovery as a focus area for the incubator. We then conducted a feasibility assessment for GPC and used our findings to provide an incubator structure recommendation as well as next steps for GPC to consider in leveraging the incubator for AI drug discovery.

Final Thoughts

Overall, our team really enjoyed working on a business-critical project for GPC. We valued the challenge of adopting a new strategic mindset (i.e., the choice structuring approach) and applying it to a real business problem. We are leaving the class more confident in our ability to approach and solve strategic questions. Thank you very much to Professors Carri Chan, Peter Tollman and Taylor Sewell for the support.

Contributors: Saba Rawjani, Sanika Chitre, Xinyi Wang, Jiying Han, Kyra Deeth-Stehlin