THANK YOU FOR SUBSCRIBING
Pharma Tech Outlook | Wednesday, August 28, 2024
Along with its enormous promise, AI poses regulatory concerns about algorithms' potential for bias, the proper use of personal data, and, most crucially, ensuring patient benefits.
Fremont, CA: In recent years, individuals bringing computer science to biology have been okay with getting funds to launch their ideas into the commercial market. The promise was straightforward: obtain massive datasets, and machine learning will tackle long-standing challenges in drug discovery, development, and testing.
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
Many of these early pharmacological programs have failed, however. The sector and patients must identify and address these increasing pains. This essay highlights three main barriers to success and demonstrates how to overcome them.
Patient Data is Fundamental
In the tech and biotech industries, using new know-how from an academic lab to address a technical challenge with well-defined datasets is a tried-and-true strategy for success. However, the application of Artificial Intelligence (AI) to medicine, sometimes known as 'techbio', is less about 'tech' and more about 'bio' in the broadest sense.
The data required for success is frequently more diverse, from more sources, and, most importantly, comes from actual patients within the complexities of numerous healthcare settings.
Too many people started with readily available materials: cell lines, chemical libraries, and mouse models. However, this undermines AI's unique potential. For the first time, we can examine truly huge and relatively new types of patient data at scale, including digital pathology slides, genomes, and proteome profiles. In other words, we no longer need to deal with simplified models of human biology.
Since it's all about the patients, the data should also be. This entails going beyond data scraping and model systems to enlist the help of the patients from whom the data is derived and the clinicians and specialists who collaborate with them.
Staying Close to Academia
Human intelligence must precede - and follow - artificial intelligence in the production of medicine. It is found at universities and research hospitals. So, forming deep and long-term alliances with universities and research hospitals is crucial to enhancing AI's success in producing better medicine.
This is not a one-way street but a network of boulevards where academics, pharma, and techs constantly interact. New means of engagement, such as privacy-preserving federated learning, enable machine learning models to be trained remotely on large multimodal data sets without physically transporting them. This can enable basic research, diagnostics, and medication discovery, providing patients with a steady stream of benefits while protecting privacy and maximizing the use of more diversified data.
Improve the Odds in Clinical Trials
The vast majority of experimental medicines fail in clinical trials. This is extremely expensive for society and industry alike. Fewer potentially promising medicines reach patients, and the expense of these failures depletes resources that could otherwise be used to fund new programs in the numerous areas of unmet need.
One of the main reasons for this is the trial's structure. In an ideal world, medications would target the underlying biology of well-defined disorders and be tested only on people who had them. Clinical trials may then be relatively modest but almost always adequate. In practice, we don't know nearly as much as we can about diseases, drug mechanisms, and trial participants' specific disease types.
More in News