3 Ways AI Can Improve the De-Risking of R&D Decisions

There are many possible roadblocks on the path to IND approval using traditional methods of drug discovery and development alone. Adverse drug-drug interactions (DDIs), ineffective drug formulation, and poorly understood patient stratification are among the top. Hurdles like these can represent “blind alleys” invisible when a candidate is initially discovered, that make carrying forward translational research and clinical trials extremely risky. Luckily, the integration of artificial intelligence (AI) into drug development can help in these three core aspects of de-risking R&D decisions and successfully getting a new drug to market.

Adverse drug reactions (ADRs) caused by unanticipated DDIs are potentially the biggest risk to a drug manufacturer because not only can these cause the drug to fail, but they can also result in patient harm. As a leading cause of patient morbidity and mortality, ADRs must be prevented to protect both the success of the drug candidate and the health of patients. These reactions are often a result of toxicity that is created when clearance functions are unintentionally blocked by the DDI; they can also be caused by linkages to/interactions with CYP450 isozymes. Fortunately, AI-enabled model-informed drug development (MIDD) methods are able to introduce a new level of accuracy in detecting and thereby preventing ADRs. These computational methods that incorporate AI and machine learning enable drug manufacturers to identify negative or unintended reactions sooner, saving time, money and other resources in the drug development process - essentially de-risking a critical class of R&D decisions in that regard. 

Furthermore, AI enables easier creation of safety toxicity profiles, which in turn aid in better prediction of DDIs/ADRs. And a major benefit of anticipating ADRs and stopping them ahead of time is enhanced efficacy of drug candidates, which is defined as a medicine’s ability to produce a desired effect or treat a specifically indicated condition. When applying for IND, patient safety and drug efficacy are both concerns that drug developers are urged to pay close attention to. Doing so can aid in de-risking of R&D decisions across the entire spectrum of pain points often encountered in the formulation process, including questions and challenges around solubility, stability, dose range and more.

Formulation - the actual building of a drug deliverable candidate through a given route of administration- is the next key aspect of drug development and discovery that can be enhanced by AI. Machine learning and artificial intelligence can assist with the entire formulation process, including determining the right combination of pharmaceutical ingredients to achieve the intended outcome, as well as evaluating a drug’s scalability and establishing the most effective treatment protocols. It can even assist with determining which route of administration the drug will work best - tablet, injection, etc. Common pain points in the formulation process include solubility, stability, dose range, all of which AI can assist with. Being able to have this level of confidence in the formulation process is important in the context of de-risking R&D decisions.

Finally, patient stratification is a critical area of drug development in which AI and machine learning can assist. The purpose of patient stratification in this context is to determine that a drug is safe for an entire population as opposed to just a select sub-group. The risk inherent in insufficient patient stratification methods is that without them, a drug developer could discover that a drug that seemed to work well in the beginning of the process does not work well in other subgroups or potentially even causes harm. Patient stratification organizes patients into different strata or ‘blocks’ according to an established criteria such as ethnicity, gender, medical history, and socioeconomic condition, and so on. An AI-powered platform that can perform analyses past the capabilities of humans doing it manually can help ensure that every patient subgroup receives exposure or allocation to experimental treatments. [Further reading: Integrating AI/ML Models for Patient Stratification Leveraging Omics Dataset and Clinical Biomarkers from COVID-19 Patients: A Promising Approach to Personalized Medicine]

In conclusion, AI is critical at this point to begin integrating into your strategy for de-risking your R&D decisions. Not sure where to start? Check out our AI adoption white paper, our ebook on the topic, or reach out to talk to a team member about the BIOiSIM platform today.