Innovation is at the core of the life sciences industry. One exciting area of growth is the increasing role artificial intelligence (AI) is playing in research and development (R&D), which today is at an inflection point as broad AI handles multimodal data systems and solves a wider range of problems. Generative AI approaches have enabled the development of these broad AI systems that are human-like and able to process vast amounts of data to generate new content and make predictions across multiple domains. One example of this is DeepMind's AlphaFold 2, which can predict the three-dimensional structures of around 200 million proteins from various organisms, helping researchers develop new therapies faster.
As companies look to the future and seek to maximize their R&D investments, many have already established data science and AI divisions. This is one of the insights we gained as part of a comprehensive, anonymized benchmarking study of 15 large and midsize biopharma companies. Our study focused on both the current state and future of R&D data science organizations, as we wanted to better understand how they are operating and evolving, while also gleaning best practices to share with the wider industry. We conducted primary interviews with key leaders in nine of the 15 organizations and spoke with ZS industry experts who work closely with these organizations. Finally, we conducted secondary research to gather insights on the six organizations we didn’t interview.
We specifically focused on five key dimensions (see figure below) that are particularly important to pharma organizations as they enhance their AI capabilities.
We found many organizations are adopting a long-term focus on AI and machine learning (ML) investments while still reaping their short-term benefits. These companies often have mandates from the C-suite—some have even dedicated leadership roles to evangelizing for AI—and are committed to making bold bets on AI-powered innovations.
Leading biopharma R&D organizations are building differentiated moats around data, expertise and community. They’re pursuing purposeful data strategies involving strategic acquisitions and partnerships, while hiring experts with knowledge of both business domains and data science. And they recognize the importance of establishing themselves as a destination for data science talent, which they do by fostering a sense of community and hosting events like hackathons, symposiums and conferences.
Anyone who has read tech news in recent months knows AI ethics are a hot topic—and they’re a key component of these organizations’ AI and ML strategies. For example, Novartis has outlined eight principles for the ethical and responsible use of AI, including empowering humanity; accountability; mitigating bias; respecting privacy; ensuring transparency; explainability; safety and security; and environmental sustainability.
The biopharma companies we observed have three primary types of operating structures for their R&D data science organizations: centralized, decentralized and hybrid. Each has its own benefits and drawbacks—the right fit often depends on the specific needs and goals of the organization.
Decisions in a centralized structure are typically driven from the top down, with the C-suite encouraging teams to integrate analytics and AI throughout the drug development process. On the other hand, decentralized models prioritize domain-specific insights and efficient service delivery. A hybrid structure, which is becoming prevalent in some leading R&D data science organizations, combines elements of both centralized and decentralized approaches.
In a hybrid structure, the central data science team acts as a hub and drives the data strategy, trains data scientists and provides new AI and self-serve capabilities to the various “spokes,” or domains, that are embedded in the R&D function. These domain-specific teams tackle custom analytics to address pressing problems, empower data-driven decision-making and work closely with key R&D decision-makers.
The role of information technology (IT) teams is also undergoing a shift in leading biopharma organizations, as IT becomes a strategic partner in the development and deployment of AI capabilities. To support this effort, many IT departments have established AI centers of excellence that bring together data scientists, data engineers and other experts to continuously improve and scale AI infrastructure and innovations.
Recent and expected rapid advances in AI, such as generative AI, have some firms assessing their future operating models. And while the companies we examined are focused on building their AI capabilities, they are also open to outsourcing or collaborating with external partners, including tech companies, academia and startups.
Complex challenges require professionals with specialized skill sets. To meet their personnel needs, leading pharma companies are adopting a T-model approach for hiring both generalists and expertise-based roles. Generalists are critical to facilitate communication and collaboration between different teams, while ensuring the data science function is aligned with the organization’s R&D goals. These professionals are often skilled at bridging the gap between data scientists and other stakeholders, such as researchers, clinicians and business analysts.
Meanwhile, those in expertise-based roles include product managers, biostatisticians, computer vision experts and more. They typically have advanced degrees in data science or specialized experience. These experts must be or become proficient in both the functional domain and data science. For example, an individual with expertise in clinical development and natural language processing could use their skills to extract valuable insights from unstructured text data like historical protocols, clinical research associate notes and health authority queries.
Beyond traditional hiring methods, firms may explore alternative approaches such as acqui-hiring, which involves continuously acquiring startups to bolster talent pools. Roche’s acquisition of Prescient Design in 2021 is one example of acqui-hiring. Companies are also exploring staff augmentation—the practice of hiring independent contractors from academia, consulting and technology companies to meet short-term needs. And some firms, as Novartis highlighted with Novartis Biome, are empowering R&D startup entrepreneurs by creating community and providing resources.
Of course, talent retention is also critical. To retain and engage their employees, leading organizations have defined clear career paths and tied performance indicators to innovation. They also offer exposure to new areas of expertise through job rotations, as a way to focus on keeping their top performers engaged and growing.
As data has become an increasingly important R&D asset, pharma companies have seen multimodal data offer insights that were previously unattainable. Multimodal data encompasses data from traditional and emerging sources such as biological signals, real-world data, images, genomic sequences, digital health devices and clinical trials. By capturing and harmonizing these data sources, researchers and scientists can gain a deeper understanding of the mechanisms of disease, identify new therapeutic targets and develop more effective and personalized treatments.
Our study indicates that leading organizations are extending their data network and tearing down internal silos between discovery, clinical development and medical affairs. Just as importantly, they’re establishing partnerships with outside firms. Tempus, for example, is assisting pharma companies in supplementing their real-world data and evidence needs by providing biopharma with a de-identified multimodal data bank, genomic sequencing products and its TIME Trial™ program, which identifies eligible patients and brings trials to them using a just-in-time activation model.
We also observed that some biopharma companies are building comprehensive data fabrics and platforms designed to enable the capture, integration and analysis of different types of data across all stages of drug discovery and development. AbbVie's data convergence initiative exemplifies this trend, as it unifies multimodal health data from various sources to expedite the discovery and development of new medicines. Companies like AbbVie are adopting FAIR principles, with the goal of ensuring findability, accessibility, interoperability and reusability in these data platforms.
It’s encouraging to see pharma companies collaborating to disrupt and improve the drug discovery process. It’s now possible for ML-driven algorithms to provide insights into data sets without sharing them, exposing them or even moving them from where they’re housed. For example, 10 pharmaceutical companies—Amgen, Astellas, AstraZeneca, Bayer, Boehringer Ingelheim, GSK, Institut de Recherches Internationales Servier, Janssen, Merck and Novartis—are contributing to the MELLODDY project, which allows these companies to safely and securely learn from each other’s proprietary data sets.
We also saw that leading pharma companies are focusing on data-centric AI, enriching data by using AI to input missing values and synthetically generate data. They are also incentivizing data investment by defining and attaching key performance indicators to data innovations.
Biopharma companies are using a variety of internal and external strategies to drive innovation in R&D. Internally, they’re building dedicated AI teams, hiring specialized talent and fostering a culture of experimentation. They’re also continuing to learn about future technologies like quantum computing, computational phenotyping and federated learning.
Looking beyond their offices and campuses, biopharma companies are partnering with academic institutions, technology companies and startups that have AI expertise and resources. Large pharma companies have agreed to more than 75 AI partnerships with institutions in recent years, including:
Partnerships with tech companies: Biopharma companies have not only partnered with Big Tech on infrastructure services, but also on R&D innovations. Calico Life Sciences, a subsidiary of Alphabet, and AbbVie are working together to discover, develop and bring to market new therapies for patients with age-related diseases, including neurodegeneration and cancer.
Partnerships with tech startups: Collaborating with technology startups gives biopharma companies access to modern technologies, approaches and talent. Biopharma companies are engaging with startups in a number of ways, including investing in them; acquiring them; co-developing new products; licensing technology or intellectual property; establishing incubator or accelerator programs; and collaborating on research projects or accessing data sets. Sanofi and Exscientia, for example, are aiming to develop up to 15 novel small molecule candidates in oncology and immunology using Exscientia’s AI-driven platform and a personalized medicine approach.
Partnerships with academia: Working with educational institutions can take many forms, including the sponsorship of research projects focused on new drug development, working together on clinical trials, resource and expertise sharing agreements and technology transfer partnerships. One example of an academia-biopharma partnership is with Janssen and MIT's Jameel Clinic, as they’re exploring how AI and ML can advance the diagnosis of disease, development of treatments, prediction of treatment response and more.
Industry consortia: It’s been positive to see biopharma companies recognize the benefits of sharing expertise and resources to address R&D challenges and opportunities. For example, MIT facilitates the Machine Learning for Pharmaceutical Discovery and Synthesis Consortium, or MLPDS, which includes 12 leading biopharma companies. The members of this consortium have a common interest in creating data science- and AI-driven algorithms and tools for discovering and synthesizing new therapeutics.
AI's potential to transform biopharma research is undeniable, but turning innovative ideas into scalable reality demands more than just a solid organization. A thorough process guiding the journey from ideation to scaling is essential. Despite enthusiasm for AI in biopharma R&D, only 20% of companies have successfully scaled and driven mainstream adoption of these innovations across their organizations.
It’s not uncommon for the buzz around a technology to exceed its promise, and AI's potential to revolutionize pharma R&D is no exception. Although AI holds great promise, only a handful of use cases— such as site selection, enrollment modeling and clinical supply optimization—have shown significant impact and gained widespread adoption.
Various barriers hinder the successful implementation of AI in R&D. These include the need for high-quality data, navigating intricate regulatory landscapes, overcoming organizational resistance to change, developing AI algorithms that are interpretable and trustworthy, and bridging the talent gap between data science and industry-specific expertise. Integrating AI demands significant workflow alterations and a workforce that possesses both data science proficiency and deep industry knowledge.
To tackle these challenges, companies are taking a strategic and purposeful approach by revamping workflows and fostering a data-driven decision-making culture. As generative AI technologies like ChatGPT gain traction, biopharma leaders must evaluate their ethical implications. Ensuring data privacy, maintaining data sensitivity and promoting responsible innovation will be paramount. By considering these factors and evolving their R&D data science teams, biopharma companies can responsibly unlock AI's potential and revolutionize the industry while reversing Eroom’s Law.