Generative AI and LLMs: Unlocking the Power of Integrative Science to Help Pharma Solve the R&D Attrition Problem

Pharmaceutical companies have an attrition problem…and it fundamentally drives the industry’s economics that dictates how long it can market a new medicine and how much it must charge for it in order to not only re-coup the cost of developing, producing, and selling that new medicine, but also re-coup the cost of all the failed drugs that never make it to market.  This attrition challenge is the economic lever that…if improved…will dictate the success or failure of major organizations across the landscape.  

Integrative science is a field of research that seeks to combine insights from different disciplines to solve complex problems. It has proven to accelerate our understanding of the physical and biological sciences and has led to new breakthroughs in many areas, including health, medicine, and materials science.   

For the development of new medicinal products, once example of integrative science take the form of human disease biology, computational research (modeling), translational research, and translational development, all working in concert to bring advances at the lab bench to the bedside of patients and back.  If the industry is ever to solve its productivity challenges – productivity losses from the failure of new drug projects and the attrition of drug candidates from its R&D pipelines in the mid to late-stages of development, then integrative science must play an even larger role in understanding causal human disease biology and integrating the understanding of the patient from the clinic back into research. 

Interdisciplinary science is the process of combining knowledge and expertise from two or more disciplines to create something new and may require years or decades before paying off. The underlying mechanism to make this happen involves integrating ideas, data, methods, tools, concepts, and theories from different disciplines. Interdisciplinary science is different from multidisciplinary science, which brings people together from different disciplines, each drawing on their disciplinary knowledge and methods so that different disciplines are integrated often in a cross-functional matrix or team within the organization. 

However, moving beyond multidisciplinary and interdisciplinary science and towards true integrative science is challenging because it requires researchers to break down the fundamental silos between disciplines and collaborate in new ways. This can be difficult in large or geographically distributed R&D organizations, as researchers often have different training and perspectives…and motivations.   

From experience, the human element of collaboration and sharing of data, insights, and knowledge is often the most difficult to overcome. 

Can Generative AI and Large Language Models (e.g., ChatGPT) help to bring together data, insights, and knowledge from across scientific disciplines that don’t normally collaborate and/or communicate within or across an organization?

Generative AI and large language models (LLMs) have the potential to overcome some of these challenges. Generative AI models can create new data and insights that can help researchers to see problems from new angles. LLMs can be used to process and synthesize large amounts of information from different sources, which can help researchers to identify common patterns and connections. 

One of the most promising aspects of generative AI and LLMs is their potential to enable integrative research. Integrative research is a holistic approach to research that seeks to understand complex systems by combining insights from different disciplines. This type of research is essential for addressing many of the world's most pressing challenges, such as climate change, disease, and global poverty. 

However, integrative research is often difficult to conduct due to the challenges of bridging disciplinary silos and integrating data from different sources. Generative AI and LLMs can help to overcome these challenges by: 

• Facilitating communication and collaboration between scientists from different disciplines. Generative AI can be used to create tools that help scientists to communicate more effectively and collaborate on research projects. For example, generative AI could be used to develop machine translation tools that allow scientists to read and understand research papers from other languages.

• Helping to integrate data from different sources. Generative AI can be used to develop tools that help scientists to integrate data from different sources into a single, unified dataset. This can make it easier to identify patterns and trends that would not be visible when looking at each dataset individually. 

• Generating new hypotheses and insights. Generative AI can be used to generate new hypotheses and insights that would not be possible to come up with using traditional methods. For example, generative AI could be used to generate new drug candidates or to identify new potential targets for climate change mitigation. 

  
  

Based upon the past 30+ years of experience bringing new digital capabilities to large Pharmaceutical R&D organizations across the Biopharmaceutical industry, I can attest that one of the great barriers to integrative, or even multidisciplinary science is the inability to find, integrate, and leverage the wealth of data that is created every day across the research enterprise.    

Here are some specific ways in which generative AI and LLMs could be used to advance integrative science: 

• Generating new hypotheses: Generative AI models can be used to generate new hypotheses about complex systems. For example, researchers could use a generative AI model to generate new hypotheses about the causes of cancer or the effects of climate change. These hypotheses could then be tested experimentally. 

• Identifying new connections: Generative AI models can be used to identify new connections between different datasets. For example, researchers could use a generative AI model to identify new connections between medical records and genetic data. This could lead to new insights into the causes and treatments of diseases. 

• Developing new tools: Generative AI models can be used to develop new tools for integrative science. For example, researchers could use a generative AI model to develop a new tool for visualizing and analyzing data from different sources. This could help researchers to identify patterns and connections that would otherwise be difficult to see. 

Connecting Data and Generative AI to Business Value and Use Cases 

Here are some specific examples of how generative AI and LLMs are already being used to advance integrative science: 

Drug Discovery

• Molecular invention: Researchers are using generative AI models to design new drugs and predict their interactions with biological systems. For example, researchers at DeepMind used a generative AI model to design a new antibiotic that is effective against a wide range of bacteria. AlphaFold has transformed the molecular invention process as computational scientists can rapidly predict the three-dimensional shape of proteins, increasing the throughput for predicting binding affinities of small and large molecule therapeutics. 

• Predicting drug properties: Generative AI can be used to predict the properties of new drug candidates, such as their solubility, toxicity, and metabolism across scientific disciplines that traditionally have not integrated their data or the science. This information can be used to select the most promising drug, and de-select ones that will likely not pan-out, in order to determine the best candidates for further synthesis, wet-lab testing, and/or development

Clinical Development 

• Designing clinical trials:  Generative AI can be used to design clinical trials that are more efficient and informative. For example, generative AI can be used to identify the optimal dose of a new drug candidate and to select the right patients for participation in a clinical trial. 

• Analyzing clinical research data: Generative AI can be used to analyze clinical trial data safety and efficacy phenotypes integrated with detailed multi-omic biomarker data more efficiently and effectively. This can help to identify the safety and efficacy of new drug candidates and to identify new insights into the disease process.

Material Science

• Materials science: Researchers are using generative AI models to interrogate the whole of the elemental table to integrate across scientific disciplines and look for new combinations of materials, designing new materials with specific properties. For example, researchers at MIT used a generative AI model to design a new composite material that is stronger than steel but lighter than aluminum. 

Generative AI and LLMs are still in their early stages of development, but they have the potential to revolutionize integrative science. By helping researchers to generate new hypotheses, identify new connections, and develop new tools, generative AI and LLMs could lead to new breakthroughs in many areas. 

Pushing right though the inevitable trough of disillusionment - challenges and opportunities for Generative AI and LLMs

While generative AI and LLMs offer great promise for integrative science, there are also some challenges that need to be addressed. One challenge is that generative AI models can be biased, reflecting the biases in the data that they are trained on. This means that it is important to carefully evaluate the outputs of generative AI models to ensure that they are not biased. 

The operating model for AI is not yet clear within large R&D organizations – Who owns AI? Who is accountable for its design and usage?  Who should ensure it meets the requirements of quality, data protection, privacy, lack of bias, etc.  In general, there is a food-fight happening in just about every major company across the globe for who owns and directs the usage of this tool, with each function staking claim to some form of ownership or accountability – business domain, IT, digital, analytics, informatics, etc. 

Another challenge is that generative AI models can be computationally expensive to train. This means that access to generative AI models is not always equitable. It is important to find ways to make generative AI models more accessible to researchers from all backgrounds.

Despite the challenges, the opportunities offered by generative AI and LLMs for integrative science are vast. Generative AI and LLMs have the potential to help us to better understand the world around us and to solve some of the most pressing problems facing humanity today.

 Are we there yet?  Not quite… 

Generative AI and LLMs are powerful new tools that have the potential to revolutionize integrative science. By helping researchers to generate new hypotheses, identify new connections, and develop new tools, generative AI and LLMs could lead to new breakthroughs in many areas. 

While there are some challenges that need to be addressed, the opportunities offered by generative AI and LLMs are vast. It is an exciting time to be involved in integrative science, and I am confident that generative AI and LLMs will play a major role in shaping the future of the field. 

Here are some additional thoughts on how generative AI and LLMs can be used to advance integrative science: 

• Finding and integrating diverse data: Generative AI can assist data scientists and data engineers to find, curate, conform, improve, and integrate data that is required for integrative science.  All too often data is locked away in fragmented silos based upon the scientific domain and transactional source data systems that were used to create the data.  There is a new era of data modernization that is working to unlock data for an multidisciplinary and integrative purposes.  Generative AI will accelerate that transformation.

• Creating interdisciplinary research teams: Generative AI can be used to create interdisciplinary research teams that bring together researchers from different fields. For example, a generative AI model could be used to generate a list of researchers whose work is relevant to a particular problem. Imagine a Gen AI algorithm providing suggestions for the best possible team composition and/or collaboration partners. This can help researchers to find new collaborators and break down the silos between disciplines. 

• Developing new educational resources: Generative AI could be used to develop new educational resources for integrative science. For example, a generative AI model could be used to create a personalized learning journey and consumer experience.  

Conclusion 

Generative AI and LLMs have the potential to revolutionize integrative science, leading to new discoveries and breakthroughs that would not be possible otherwise. By addressing the challenges of using these technologies, we can usher in a new era of scientific discovery and innovation – and finally tackle the many challenges to the probability of success for a new medicine as it winds it’s way through the R&D process.  There is a race to understand and leveraging new AI capabilities to tackle the challenges of precision discovery and development, and improve R&D productivity by reducing the attrition problem. 

At XponentL Data, we are working with leading organizations to unlock the power and value of their data as they take on the challenges and opportunities presented by the new era of Generative AI.   

 Reach out today and contact us to start a conversation on how we can help your efforts.