When can we expect AI-developed pharmaceuticals?

Artificial intelligence (AI) has become a powerful tool in many industries, including healthcare due to its long and close links with medical research. Indeed, studies on the ways that human neurons interact formed the basis of the first artificial neural networks that ultimately led to the AI models we see today.

Since the 1970s, the possible applications of AI in the medical field have mushroomed, leading to faster and more accurate diagnostics, and helping to relieve medical staff of routine tasks. The growth of AI in the healthcare arena is set to continue well into the future, with estimates of market volume increases that range from 25 % to 35 % or more a year up to 2035, depending on the data used to calculate the figures.

A natural partner for drug discovery?

While AI use is different in every sector of every industry, AI in drug development has not achieved its full potential. So far, not a single drug has been developed - from initial idea to market approval - using AI alone. At first glance, this appears rather surprising. After all, AI capabilities seem a natural partner for pharmaceutical research and development.

In part this is true. Even today, AI plays a significant role in supporting the development of new drugs. Overall, though, while AI currently complements the process, it cannot yet carry out every stage of drug development.

Why is this? The problem is multi-faceted. To begin with, human biology is extremely complex and often not fully understood. This level of uncertainty hampers AI from optimally performing the task of analysing and mapping the relationships necessary to drive drug development. 

Fragmented data, impenetrable AI models

In addition, there is no central global database that holds all the available clinical and biological data. Even if there were, the data probably couldn't provide sufficient detail and context. Rather, information is held in myriad databases, which are sometimes inaccessible.

Even when AI hypotheses or potential active substances are identified as having drug development potential, they need to be validated through thorough laboratory studies and clinical trials. The strict regulatory requirements for drug trials and testing also demand comprehensive evidence of safety and efficacy, which AI-generated results cannot (yet) fully provide.

AI models themselves also come in for criticism from both regulators and medical researchers. These models are highly complex, which is why they are often called "black boxes". Their lack of transparency makes them difficult to check and interpret, leading to scepticism about their results.

Humans still required

The problems with drug development demonstrate the limits of AI today. Without further AI progress, the pharma industry remains reliant on human intelligence in the process. 

So what can AI contribute in the pharma sphere? Plenty. AI can help locate potential future drugs through the analysis of large molecule- and other databases; support the early identification of possible toxicity, side effects, and adverse reactions; optimise study designs by selecting suitable candidates for inclusion; speed up the analysis of clinical trial data; and assist with the creation of personalised therapies based on genetic data.

AI can also play a role - as it has in many industries, not just in healthcare - in reducing both the time and money needed for new product development. While the creation of a new drug using traditional methods costs an average of USD 2 billion to USD 3 billion and takes ten to 15 years, current estimates suggest that the use of AI could cut both costs and development time by an average of 30 % or more.

Speeding up drug design

One of the first pharma product candidates supported using AI was DSP-1181, a selective serotonin receptor agonist (SSRA) used for the treatment of obsessive-compulsive disorder. With the help of AI and automated experimentation, the active ingredient was designed in just 12 months, rather than the several years this would have taken using traditional methods.

Although the drug was discontinued at an early stage (in 2022), it is considered a milestone and proof of concept for AI drug-development usage, as it was the first AI-based molecule to reach the clinical trial phase.

Though the future of the pharma industry undoubtedly includes AI, it remains to be seen to what extent it can take drug development out of the hands of human scientists and researchers.