How AI can clean up the  pharmaceutical industry

In the pharmaceutical industry, companies spend several decades, tens of billions of dollars, and countless hours experimenting on animals and humans trying to prove a hypothesis that a certain disease mechanism can be addressed with a specific class of drugs.

Ninety per cent of these endeavours fail in clinical trials and even more fail in pre-clinical research, leaving behind vast quantities of what is known as “dark data” – or data that has been collected as part of research and development that has not been processed or analyzed.

Very often, drugs that fail in their intended use may work elsewhere. One famous example is Viagra, which was tested for a certain cardiovascular condition before becoming a blockbuster erectile dysfunction drug. The current headline-generating weight loss drug, Ozempic, got its start as a diabetes medication.

Most drugs, though, do not get a second chance. That creates a vast wasteland of dark data, research and funding. This massive waste makes pharma among the most inefficient industries on the planet. If we’re going to fix pharma’s inefficiencies, it’s critical to first examine the drivers of this waste. The list of reasons pharma terminates programmes is very long and includes:

Often, the drug does not work to treat the condition for which it was designed or it is too toxic. Human biology is complex and may not be possible to find a good indication for a drug in development.

Patents provide pharma companies with an exclusive 20-year span when they have a monopoly on their drug and can set prices. But even a 20-year patent may not be enough. On average, it takes 10 years for a drug to be approved. If it takes longer due to company handovers, minor failures, regulatory delays or difficulty raising funding, it may not be economical to proceed. The newly introduced Inflation Reduction Act (IRA) makes it even more difficult as many pharma companies are deprioritizing the development of small molecule drugs in favour of very expensive biologics (products derived from living organisms, such as vaccines), which offer additional ways to protect their business.

Imagine a hypothetical scenario, where there is 10 years left on a patent and a company fails in a Phase II clinical trial for prostate cancer, but discovers that the same drug may work for breast cancer. After re-doing the work for the new indication, it would take them another six to eight years to get the drug approved – leaving just two years of patent protection before this drug can be made and sold by other companies.

During my 10 years as CEO of Insilico Medicine, I’ve seen several CSOs and heads of pharma R&D change their jobs multiple times. Many pharma companies have changed their CEOs at least once. Every time there is a change in management, changes in research and development follow. The new bosses may cut or increase the internal R&D, but they almost always shift the focus from one industry to another. I once saw a new CSO come in and eliminate close to 70% of the internal R&D and most of the external R&D efforts, shift the company’s focus, strike big-ticket deals with close colleagues, and leave in just a few years for a company founded by one of these colleagues.

When a hot new technology emerges, pharma companies often chase that opportunity. The recent progress in anti-obesity drugs pioneered by Novo Nordisk and Eli Lilly caused many pharma companies to cut entire therapeutic areas and departments to refocus on this emerging field. Many understood that they missed out on the specific approach to injectable biologics and started buying oral small molecule inhibitors. Others realized that some of these anti-obesity drugs have potential liabilities including muscle loss and jumped into anti-muscle wasting medications. Many promising cancer drugs are likely to be abandoned in this fundamental shift.

Since there are thousands of people working over many years on a therapeutic programme before it launches, it is difficult to assign blame to a specific person or a small group of people when it fails. Very rarely can one CEO alone be credited with prioritizing the early-stage research platform, discovering a drug, completing the clinical trials and putting the drug on the market.

The best way to recycle is to prevent the creation of waste in the first place. Generative AI has the ability to dramatically increase the chances of success of a therapeutic programme by helping to pick the right disease target and generating a highly optimized molecule with desired properties, instead of searching for a needle in a haystack and identifying the list of indications where the target/mechanism-drug pair is likely to work.

Pharma companies that truly embrace AI could run parallel clinical trials for the same drug as a single agent or in combinations targeting multiple diseases where the drug is predicted to work. This approach may help address several diseases simultaneously and drive profits while the drug is still on patent.

Generative AI systems can also track company leadership and researchers responsible for key decisions from program initiation to completion. Generative AI platforms trained on pharma data can be initiated as multi-agent systems to help hire and track the performance of pharmaceutical executives, assist with objective-oriented business development, licensing, and acquisitions, and provide real-time reporting on results and progress.

Pharma companies are facing massive losses of exclusivity, with 20-70% of their blockbuster therapeutics going off patent and limited opportunities for potential blockbuster acquisitions. These companies need to start placing big bets with accountability and real-time reporting on both internal R&D and external partnerships. These decisions can be made much better with the use of the latest generative AI platforms.

Of course, these changes are not easy to implement because investors, employees, board members and even policy-makers tend to resist substantial changes. Recently, the new CEO of a big pharma company put a plan in motion to revamp the internal research to replenish the pipeline with internally developed therapeutics and early-stage licensing deals, and the company’s stock fell 20% in a single day.

In the short term, we should expect the market to punish these bold but necessary moves until the first companies adopting all-in generative AI strategies succeed. This will require several years. But in the end, accelerating the adoption of generative AI in the pharma industry is vital not only to the reduction of wasted resources, but to extending quality of life for everyone.