Over the course of the past two years, NewtonX has conducted a series of one-on-one consultations with senior researchers at Berg, Johnson & Johnson, Pfizer, and several other pharma companies about the costs and processes associated with drug discovery. Through these interviews, we have drawn numerous conclusions about the potential for AI in drug discovery at big pharma companies.
According both to the NewtonX researchers interviewed for this piece, as well as third party studies, the cost to develop a new pharmaceutical drug exceeds $2.5B, a number which has more than doubled in the past 10 years, even accounting for inflation. To make matters worse, a significant portion of this money spent is effectively wasted, because it includes money spent on the nine out of 10 candidate drugs that fail at some point between phase I trials and regulatory approval. This rise in costs comes despite serious efforts to reduce R&D costs by big pharma companies, but because of increased complexity in clinical trials, a greater focus on comparative drug effectiveness data, and increased investment in chronic and degenerative diseases, the efforts have proved ineffective.
Because of this, Big Pharma is interested in the possibilities of using AI and Machine Learning to improve efficiency, speed, and accuracy in drug development research. In fact, as early as 2007, a robot called Adam identified the function of a yeast gene by searching public databases, deriving hypotheses based on the data, and testing the hypotheses in the lab with a robot. Because AI is adroit at combing through mountains of data, identifying patterns, and testing conclusions drawn from these patterns, it can significantly help R&D processes.
Cancer Treatment, Drug Validation, and Clinical Trials: AI Can Do It All
A group of researchers at Berg, a pharma company based in Boston, applied Machine Learning to discover a new drug, now in phase II clinical trials, for treatment of advanced pancreatic cancer. The company uses a proprietary platform that identifies biomarkers by applying an AI-powered algorithm to analyze large numbers of patients’ genotypic and phenotypic characteristics. Berg was able to create the first complete model for how pancreatic cancer functions, and also pinpointed exactly what allows cancer to grow. By identifying prominent factors in how pancreatic cancer develops and spreads, Berg was able to develop a cancer drug, dubbed BPM 31510. The drug has the potential to cure the third-most-common cause of cancer death in the U.S., and save the 53,000 people who are diagnosed with pancreatic cancer each year (73% of whom typically die within a year of diagnosis).
By eliminating the trial and error approach of the past, pharma companies can derive more precise hypotheses, and test them more rapidly. For instance, a company called Atom uses AI to predict how molecules will behave in the body, thereby allowing researchers to accelerate development of more effective therapies. Similarly, Reverie Labs predicts potency of small molecules, and conceives new molecules to optimize for them using AI, which also accelerates drug development by generating new molecule leads.
AI is also being used to maximize efficiency and adherence to protocol in clinical trials. For instance, a company called AI Cure visually confirms medication ingestion through smartphones, which helps improve adherence and validation in trials. Other companies, such as Brite Health, use AI to analyze structured and unstructured clinical trial data to reduce dropout rates. There are also biomarker monitoring platforms and predictive analytics platforms that both help improve and speed up clinical trials.
A New Opportunity for Medical AI Developers
AI has the potential to reduce the number of drugs that don’t make it past the clinical trial stage, while also speeding up the timeframe for getting to that stage. This holds massive economic potential: there are already hundreds of startups devoted to using AI for improve R&D in pharmaceuticals. Big Pharma has also invested millions in developing new AI and Machine Learning-driven techniques for various stages of the drug development process — from understanding the mechanisms of a disease, to designing drugs and running preclinical experiments.
The future of biomedical research will demand people skilled in AI and coding who can apply algorithms to the process of drug discovery. Cutting costs and improving efficiency to cut down the time to clinical trials will be paramount, and will open up a new set of doors for AI developers, researchers, and pharma companies alike.