Pharma Companies Use AI to Advance Drug Development

The biopharmaceutical industry has launched numerous initiatives on how to leverage artificial intelligence (AI) to advance drug development. For example, Pfizer expressed how AI can be used in drug discovery. Emerging biopharmaceutical companies (EBPs) are getting in on the AI action by leveraging AI to discover novel applications for shelved therapies that never made it to market. In this interview, Vimal Mehta, CEO, and Vince O’Neill, CMO, at BioXcel Therapeutics will discuss how they are actively using AI to discover advanced therapies in CNS and oncology.

Moe Alsumidaie: Why is using AI and machine learning in drug development so unique, and how does it make an impact in the future of drug development?

Vimal Mehta

Vimal Mehta: I think information is growing at a much faster rate than in the past. And particularly in drug development, information flow is even more complex because it requires solid domain knowledge of science, the clinical side, the commercial side, the IP landscape and, the competition. It’s difficult for anybody to keep up with 5,000 publications that are coming out on a daily basis and integrate and assimilate that information in their decision process. So, machine learning technology and AI is at a point where these technologies are matured, and they can be applied. That's precisely what BioXcel Therapeutics is doing.

We use AI across the development spectrum. First, in the identification of novel drugs. And once we identify them, we use it to design our preclinical and translational work. If all of that has gone well, then in the clinic, we continue to use these technologies for the identification of the patient. So, it's across the spectrum and in the future we believe that we will be able to utilize AI in real-world data. In addition, we use AI in combination with wearable technologies  with our therapy, BXCL501; It’s our intention that the wearable will alert the patient to the onset of an agitated episode, and potentially avoid it. 

Alex Neumeister: Do you have any suggestions if certain diseases or disease entities are more suitable for this type of approach or you think you can apply this across the entire spectrum of medical illness?

Vince O’Neill: These approaches can be beneficial especially in oncology; it's not that long ago, the oncology treatment we used to use was sequential monotherapy, or in other words, we would give one drug, it fails, we switched to another drug, it fails, and then we try a third one. These days we've learned a lot from infectious disease and other therapeutic areas, and we know that combinations of drugs and cocktails are really the way to go. I believe AI has a lot to offer there, allowing us to combine drugs in ways that match the known targets. 

MA: How is your approach improving the chances of success in CNS disorders? 

VM: At BioXcel Therapeutics, we decided that we're going to focus on treating neuro symptoms rather than doing the disease-modifying agents. If we identify new targets for a disease-modifying agent, we park it, and we put it into our partnering bucket because we don't have the required capabilities to develop that drug. We develop therapeutics focused on drugs that have unexploited potential that have gone to the clinic with positive phase I data; however, never made it to market. We believe that 3,000 drugs exist in that state, and if we can even succeed with only 1% of them in identification through AI, that could lead to 30 NDAs for us. And BXCL501 is a typical example of that. 

BXCL501 is used as a sedative anesthetic in a surgical unit, and we are using our technology that identified the target of agitation. This is where AI support has helped us in an area we would not have been able to achieve on our own. We expect to complete our schizophrenia/bipolar phase III trial in the first half of 2020 and file our first NDA in the second half of 2020. We did a lot of preclinical translational and human proof of concept work using the current IV form of the drug before we started testing it in schizophrenia patients. We saw significant responses with multiple doses up to a 90% response rate. 

MA: Other companies have tried to find meaningful treatments for pancreatic cancer. Yet only less than 10 drugs have been approved. How are your approaches with AI changing the realm of hard-to-treat indications?

VO: We have now a better and more in-depth understanding of cancer at the biology level. I think that we can really leverage the AI approach to rationally seek targets and combine them with drugs and also combine different cocktails of drugs. I think that's where AI really is beginning to prove itself. Pancreatic cancer is challenging to treat. From my point of view, I think that justifies a three-drug combination. And again, trying to move away from the nonspecific chemotherapies because they're limited in their efficacy. This really is the poster child for what AI can do within drug development in oncology, not only for pancreatic cancer but, also for different pathways that are unique for other types of cancer, such as lung and colon cancers.

Alex Neumeister is Head of Medical Affairs at CliniBiz and specializes in protocol design, drug safety and clinical trial management. Moe Alsumidaie, MBA, MSF, is a thought leader and expert in the application of business analytics toward clinical trials, and Editorial Advisory Board member for and regular contributor to Applied Clinical Trials.