Comment: How AI can help the pharmaceutical industry

Healthcare Today

Source published: Site published:

Original article · فارسی

By Damien Ng
How AI can help the pharmaceutical industry

Damien Ng, senior research analyst at Julius Baer, says AI cannot guarantee that new drugs will succeed. But it could help researchers spot possible failures earlier.

The pharmaceutical industry celebrates big success stories. A new treatment reaches patients. A disease that once threatened lives can now be treated. But people pay much less attention to the many ideas that fail.

For every medicine that succeeds, thousands of possible drugs are dropped during development. Years of research can end with a difficult discovery: the molecule was wrong, or the original idea about how the body works was wrong. Only about one in ten drug candidates that enter clinical trials eventually gets approved. The process can take more than ten years and cost billions of dollars.

One of the biggest challenges in this industry is working out which scientific ideas are most likely to succeed. This is why AI could have its greatest effect on the development of new medicines, not by replacing scientists or running laboratories automatically, but by helping researchers make better choices before spending years of work and huge amounts of money. Developing a drug is largely about predicting what will happen.

Faster isn't always better

Much of the discussion about AI focuses on speed. People talk about finding drugs and designing molecules faster. These benefits are real, but they may overlook a bigger point: getting to the wrong destination faster is not very useful.

Scientists need to find out whether a biological target plays an important part in a disease. They also need to understand how a possible drug molecule will interact with that target and whether the treatment can make a real difference to patients. Every choice involves some uncertainty.

For many years, researchers have used their scientific knowledge and experiments to deal with these challenges. Today, they can also use much more biological data than ever before. Information about genes, proteins, medical images and patients' real-world health records gives scientists a much fuller picture of human health. But no team of people can fully understand data on such a large and complex scale.

This is where AI can make a practical difference. Computer programs can examine millions of pieces of data to find useful patterns. Research teams are using AI models to find links between genes and the biological processes involved in disease that were previously hidden in large datasets. For example, DeepMind's AlphaFold predicted the structures of more than 200 million proteins. This gives scientists a powerful resource for understanding the biology of diseases and finding possible targets for new drugs.

AI is also starting to play a role in the business of developing drugs. Companies such as Insilico Medicine have moved drug candidates developed with AI support into clinical trials. This shows how computer-based methods can help find and choose promising treatment ideas.

AI could improve the way researchers choose drug targets. This matters because many drug research and development failures begin with decisions made at the very start. Choosing the wrong target can lead to failure years later, after a great deal of money has already been spent.

Anyone who has worked in pharmaceutical research and development knows that scientific setbacks are a normal part of the process. The goal should therefore be to spot failures early, not to remove every possible failure. AI can help researchers identify weak ideas about how biology works before they enter costly laboratory and clinical programmes. This does more than save money. It allows resources to be moved towards ideas with a better chance of success, making the whole process of developing new medicines more productive.

AI can also find useful patterns in clinical trials. Groups of patients are rarely all the same. Two people with the same diagnosis may have very different biology. AI can help researchers spot these differences, choose more suitable patients for trials and get clearer evidence about which treatments work and which patients benefit from them.

Damien Ng, senior research analyst at Julius Baer.
Damien Ng, senior research analyst at Julius Baer.

Changing what gives companies an advantage

The growing use of AI in drug research and development could also change what makes pharmaceutical companies successful. AI may make some parts of drug discovery available to more companies by giving them access to advanced tools for studying data. But it could also make unique biological and clinical datasets more valuable. Computer programs can often be copied or recreated, but private patient records and decades of research cannot. As AI tools become more widely available, access to special and hard-to-copy data may become an even bigger source of competitive advantage.

Companies with unique datasets and deep knowledge of diseases are building advantages that may be hard for others to copy. The next successful companies in the pharmaceutical industry will be those that turn data into useful knowledge. Better predictions can then help lead to better medicines.

This does not mean that computer programs will replace researchers. Scientific breakthroughs still need human curiosity and good judgement. Biology is too complex, and healthcare is too important, for people to trust computer-generated results without question.

AI's real value is much more practical. In an industry where a single decision can cost hundreds of millions of dollars, even small improvements in predicting success and failure can have major effects. The future of drug development may depend on making better choices earlier and finding out which ideas are worth pursuing before years of time and money are spent.

Developing new medicines will always involve uncertainty. AI may improve the chances of success by helping researchers make better-informed choices about which ideas to follow, but it cannot guarantee that a drug will succeed.