The life sciences sector puts years of investment and work into discovering new drugs, and the payoff comes when those drugs can be used to save lives or radically improve patient treatments. While breaking down molecules and analyzing chemical combinations takes time, the most tedious part of this critical process may lie in the production of clinical studies and reports and the regulatory red tape that can prolong the effort.
Developing studies, analyzing the results of them, writing up clinical reports, and editing and revising them to address pertinent questions and satisfy legal and regulatory requirements can be a major impediment to the entire process, according to Emmanuel Walckenaer, CEO at Yseop, a French company the develops technology around natural language generation and Generative AI. AI is being brought unto many industries to increase automation, introduce more powerful analytical capabilities, and speed productivity, and while it can be applied directly to the scientific aspects of the drug discovery process, it also can be used to help speed up and improve the accuracy of studies and reports, regulatory applications, and legal filings.
“Pharmaceutical companies are against the clock when it comes to bringing lifesaving drugs to market,” Walckenaer told Fierce Electronics. “The process is rigorous and consists of discovery, pre-clinical trials, clinical trials, clinical development and government approval. Implementing Generative AI solutions can significantly expedite previously tedious processes, while ensuring accuracy at scale. There are countless ways Generative AI can be used to automatically generate content, including articles, pictures, stories, social media posts, and more. This also has an impact on the drug discovery process in a variety of ways including new molecule creation, supporting bio-statisticians in their clinical trial data analysis, and even helping scientific writers complete the required documents for the FDA.”
Regarding that tedious documentation process, Walckenaer added that using Generative AI can give scientific teams more time for the actual science involved in chemical analysis and drug discovery. “Generative AI can certainly help discover new drugs with all of the time that is being saved by using the technology to write clinical study reports,” he said. “These reports often take up weeks of scientific writers’ time and require a thorough process to ensure that what will be submitted to the regulators is accurate. This process historically bottlenecks the advancement of studies and time it takes to get drugs to market. With Generative AI helping cut days of work out and providing complete accuracy, this opens up more time to discover new drugs that might have not been thought of before.”
However, do not expect public Generative AI offerings like ChatGPT to play a role in this process, as Walckenaer noted that companies in regulated industries like the pharmaceutical sector need to maintain confidentiality. “The [Generative AI] solutions must come from a closed source, which is proprietary and only available confidentially to the company working with it. Hybrid models are the key to success here, for example, an application that blends the power of large language models with customer data in a closed, secure environment.”
That reality has paved the way for Yseop to provide its proprietary NLG technology to major pharma firms like Eli Lilly and Sanofi, both of whom the company has engaged with over the last year.
Walckenaer described Yseop’s work with Lilly: “Eli Lilly partnered with Yseop to implement an automated process for patient narratives. With the automation of patient narratives, writing teams saw immediate value and could focus on more complex, scientific work. Before working with Yseop, one patient narrative required an average of four hours of data consolidation and cross-functional writing. Lilly’s process was manual and resource intensive, which consisted of tedious quality checking and copying and pasting. Now, the process, which took four hours for simple narratives, takes a mere four seconds and generates hundreds of them.”
As Generative AI finds a place in the life sciences sector, Walckenaer sees the potential for it to be integrated with other emerging technologies, such as quantum computing, which on its own also is being explored by life sciences companies for its ability to provide fast and accurate simulations of chemical interactions, and powerful data processing.
“Quantum computing and generative AI are a good pairing because you can use quantum computing to help train generative AI models much faster than normally,” he said. “These models consume tons of time and data and quantum computing can help train more models using less time and less data.”