AI-Powered Drug Discovery: Identifying Safety Red Flags
Unmanageable toxicity accounts for 30% of clinical drug development failures and can cause severe side effects and potential harm to patients. Download our report to see how AI-powered drug discovery can help mitigate late-stage clinical failures and market withdrawals.
The High Stakes of Toxicity
Unmanageable toxicity accounts for around 30% of clinical drug development failures and can cause severe side effects and potential harm to patients.¹ Even after rigorous testing for safety in clinical trials, a drug’s full safety profile often only emerges after extended clinical use.² Consequently, this can prompt market withdrawals, leading to substantial financial losses for pharmaceutical companies.
One major contributor to failures due to safety concerns is a poor understanding of drug interactions with targets and pathways. In fact, 75% of safety closures during preclinical testing were due to the compound being off-target.³ A comprehensive understanding of relevant pathways and targets early on the development pipeline is therefore key to anticipating potential side effects.⁴
From Trial-And-Error to Data-Driven Strategies
With the number of biomedical research articles increasing by 2.5M+ publications each year,⁵ identifying relevant disease insights is akin to finding a needle in a haystack. The ability to analyze scientific data accurately and at scale therefore confers advantages in drug development. As such, the pharmaceutical sector is experiencing a paradigm shift from trial-and-error approaches towards more rational, data-driven strategies.⁶ Notably, pharma companies are adopting AI models into their pipelines to shorten R&D cycle times and reduce costs.⁷ In this blog, we explore how AI-powered drug discovery could help preven safety-related drug attrition.
AI-Powered Insights: Anticipating Safety Issues
AI excels in the rapid analysis of large quantities of data, empowering researchers to make better-informed decisions in a fraction of the time traditionally required. AI-powered drug discovery can accelerate preclinical safety and toxicity assessments by enhancing our understanding of underlying disease mechanisms, in addition to predicting the potential side effects linked to target modulation.
Case Study: The Safety-Related Withdrawal of Umbralisib
Background: In 2022, the FDA withdrew its approval for umbralisib (used to treat marginal zone lymphoma and follicular lymphoma) due to safety concerns relating to increased risk of death.⁸
Targets of Umbrasilib: 18 genes and proteins targeted by umbralisib were uncovered by Causaly, enabling the identification of potential off-target effects. Phosphoinositide 3-kinase (PIK3CB) and casein kinase I (CK1) had the strongest evidence as targets for this drug.
Side Effects: Causaly uncovered 450+ potential side effects of PIK3CB dating back to the 1990s, including myocardial dysfunction and liver injury. Fewer were identified for CK1, with around 40 possible side effects.
Diseases Affected: Almost 750 diseases reported to be affected by PIK3CB inhibition were identified by Causaly, including hyperglycemia and various cancers. In comparison, around 80 diseases affected by CK1 inhibition were uncovered from the biomedical literature.
A comprehensive understanding of targets, pathways and mechanisms is crucial for identifying potential safety red flags early on in the development process. By rapidly processing large volumes of data into actionable insights, AI can help forecast potential toxicity concerns enabling the prioritization of more promising clinical drug programs. The use of transparent, user-friendly AI platforms will become increasingly important for preventing safety-related clinical failures and market withdrawals, thereby accelerating the delivery of safe and effective treatments to patients.
More on Safety