Harnessing Artificial Intelligence in Drug Discovery and Development

The integration of artificial intelligence (AI) into drug discovery and development endeavors has the potential to revolutionize the anticancer pharmaceutical landscape and unlock unprecedented efficiency, precision, and innovation. By leveraging machine learning algorithms, natural language processing, and data analytics, AI accelerates the identification of potential drug candidates, optimizes preclinical and clinical testing, and reduces costs. A recent study demonstrated that AI-discovered drugs in phase 1 clinical trials have a better success rate compared to traditionally discovered drugs, with estimates ranging from 80% to 90% for AI-developed drugs versus 40% to 65% for drugs discovered via traditional methods. Notably, nearly 30% of all AI uses I for drug discovery and development are focused on anticancer drugs.

In June 2024, The New York Academy of Sciences and Cure partnered to host an expert panel discussion titled The Science and Business of AI-Driven Drug Discovery. Speakers at the event explored ways that AI can expedite target identification, predict compound interactions, and optimize clinical trial design. Through this lens, I will examine the transformative role of AI in drug development, anticipated challenges, and potential future applications of this technology.

Challenges of Traditional Drug Discovery

Traditional drug discovery is a complex, time-intensive, and costly undertaking. To bring a single drug to market, efforts typically span over a decade and incur an average cost exceeding $2 billion. Each stage—from target identification and validation to preclinical testing and clinical trials­—is marked by immense trial-and-error experimentation with huge sunk costs along the way. Attrition rates are very high, with nearly 90% of drug candidates failing due to insufficient efficacy or unforeseen safety concerns.

Applications of AI in Drug Discovery

• Target identification and validation

  AI algorithms have the power to analyze complex biological datasets and uncover disease-causing targets (eg, proteins or genes). Maria Luisa Pineda, co-founder and CEO of Envisagenics, explained, “Genome analysis used to take 8 to 10 weeks per genome, whereas now, high-performance computing allows for 2000 patients’ full transcriptomes to be analyzed within 2 hours.” Machine learning models can subsequently predict the interaction between these targets and potential drug candidates, streamlining the process of target validation. For instance, DeepMind’s AlphaFold protein structure database has already revolutionized protein structure prediction, offering invaluable insights for therapeutic discovery.

• Predictive toxicology and clinical trial design

  AI models predict the safety profile of drug candidates by analyzing preclinical data, minimizing the risk of adverse events during clinical trials in humans. Machine learning algorithms can identify toxicological patterns such as hepatotoxicity or cardiotoxicity, enabling early-stage elimination of drug candidates that lack sufficient safety. AI can also be used to optimize clinical trial protocols, predict outcomes, and stratify patients, ensuring inclusion of individuals most likely to benefit from a treatment. Grant Mitchell, co-founder and CEO of Every Cure, encapsulated the vision for AI’s future, stating, “In 100 years, we’ll look back and say, ‘I can’t believe we actually used to test drugs on humans!’”

• Drug repurposing
  Another useful application of AI is the repurposing of existing drugs for new indications. By analyzing vast datasets of clinical and molecular data, AI reveals unexpected efficacies of certain drugs against unrelated diseases. Mitchell emphasized this potential, noting how AI could uncover a “useful article in a lower impact journal that no one has ever seen, but [reveals that a drug has] already worked in a human!" By breaking down silos, AI can connect the dots between scientific discoveries that may otherwise go unnoticed.

Anticipated Challenges

Despite its immense promise, the adoption of AI in drug development is not without unique challenges. Machine learning models require high-quality, diverse datasets for training and validation, but inconsistent or incomplete data can compromise model accuracy and external validity. The black box nature of many AI algorithms also makes it difficult for scientists to interpret predictions, raising concerns about reliability and accountability for critical decisions. Moreover, regulatory agencies are still adapting to the use of AI for drug development. Establishing guidelines and best practices for AI validation and ethical use of patient data remain ongoing challenges.

The Future of AI in Drug Development

As AI technology continues to evolve, its role in drug discovery is expected to expand. Emerging tools such as quantum computing could further enhance AI’s computational capabilities, enabling faster and more precise predictions. Additionally, the incorporation of multi-omics data (eg, genomics, proteomics, and metabolomics) will provide further insight into disease mechanisms, target identification, and drug design. Collaboration between pharmaceutical companies, AI startups, academic institutions, and policymakers will be essential for fostering an ecosystem in which AI-powered drug discovery becomes the standard.

Nicole A Colwell, MD, is a senior editor/medical writer for the Association of Cancer Care Centers (ACCC).