Why pharma is the next frontier for AI

Smaller healthtech firms can thrive alongside big pharma, thanks to rapid progress in the application of AI, says leading patent attorney Peter Finnie.

The pharmaceutical industry has been dominated by large pharmaceutical companies, often known as “big pharma”. This was for a very good reason. Developing drugs is incredibly expensive, time-consuming, and risky. Pharmaceutical companies spend hundreds of millions of dollars and years discovering new drugs, testing them, and then seeking regulatory approval. However, the majority of promising drug candidates fail to obtain regulatory approval because they do not have the necessary level of clinical benefit or have unacceptable side-effects. Artificial intelligence (AI) is changing the landscape by shortening discovery times whilst reducing the number of failed drug candidates.


In recent years, AI has become ubiquitous with modern businesses. Far from the realms of science fiction, almost every sector and industry has been changed in some way using AI to automate previously manual processes that took humans far longer to carry out. From finance to agriculture, AI has been implemented to assist humans in their work, improving accuracy, decision making, and time efficiency.

The healthcare and especially healthtech industries are no different. Previously, healthtech companies developed traditional software technology to remind patients to take pills, facilitate virtual doctor’s appointments or allow those with diabetes to track blood sugar levels. Although these software applications are entirely useful, AI has now swept in and provided an entirely new and exciting opportunity for healthtech companies to interact with the pharma pipeline. Most importantly, the computing power of AI algorithms has specifically impacted the way healthtech companies can now enter the lucrative drug discovery, drug repurposing, and personalised medicine markets.

The growth of AI healthtech startups has given rise to a need for patenting of not just the computer software but also inventions derived using the software to protect startups from losing out on monetising their innovations. However, using AI to help facilitate invention or innovation has become a contentious issue in recent months with the DABUS AI inventor patent cases receiving media attention on the issue as to whether an AI platform can be named as an inventor in a patent application – the answer was a firm “No”! The important thing to note is that in most cases in healthtech AI is not actually inventing but rather facilitating and speeding up innovation. There is no question that you can patent the insights that AI provides.

The high barrier of entry to the pharma pipeline has been broken down by the introduction of AI that can do much of the leg-work operating on huge data sets using the power of modern computer processors, and at a fraction of the cost. What previously took the likes of AstraZeneca and GlaxoSmithKline thousands of iterations using hundreds of pharmacists and lab hours can now be done by a handful of data scientists and pharmacists with a computer and access to appropriate data sets. The ability to patent computer assisted discoveries allows AI startups in this field to quickly and securely monetise them to allow the company to become revenue generating.


AI has allowed these startups to process vast amounts of patient data and drug data to find new drug treatments. For example, AI can be used to design the ideal structure for a completely new drug, by crunching data regarding the biological target. Al can also be used to match a disease with an unmet need with already-approved drugs, by analysing the complex pharmacology of drugs and the physiology of a disease. As every drug and disease has a profile, the computer can match the disease with a possible treatment. What the computer can do is match these elements rapidly and without stopping, whilst possibly learning which criteria are the most important. The in silico data that AI provides may not necessarily yield new drug candidates, but there is no doubt it aids the drug discovery process by narrowing down the possible candidates and thus reducing the workload for the pharmacologists. It is an important tool.

The drug candidates that may be identified by AI still require real world testing, but the time to reach this point is shortened. Once the drug candidate has been identified and verified in the lab, patent applications can be filed in the usual way. This combination of real-world data and a patent application has significant value and can be taken to a large pharmaceutical company for partnering, for example. Big pharma are often best placed to finance the large scale clinical trials needed before a drug can be approved.

By using this strategy, both the tech startups and the big pharma “win”. The tech startup is able to deliver a partnerable asset in a realistic timescale (that often ties in with the investors’ requirements) and the big pharma saves money and time that they would have otherwise have needed to spend in early stage research (which for big pharma can be very costly due to the methods they use).

Entry for tech startups funded by venture capital to do drug discovery using AI is now far lower. Previously companies were having to raise millions of pounds just to get to the stage where it had a potential drug candidate. Investors faced the prospect of putting in large sums of money and gambling that an effective drug was found. Often this didn’t happen, and the investors would lose everything. Now with the use of AI, investors can fund a startup business with a much lower level of capital and with increased confidence that the technology is going to deliver effective solutions.

 These new technologies are also applicable to vaccine development. Traditionally, vaccine development is very slow and very difficult, especially for certain viruses. Despite this, AI is still being trialled in the search for vaccines, with some early success being shown.


The key with AI is that the name somewhat misconstrues what it actually is. At present, AI is a complex algorithm or set of algorithms that churn through vast amounts of data to provide outcomes or insights. It is a tool. It does not answer a question, because it does not know what the question is. It does not invent. It assists pharmacists and data scientists in faster innovation to make discoveries.

It is important to train the machine on reliable data and this is why it is vital that data scientists are involved in training the algorithms on good, unbiased data. Large medical research institutions, including the NHS, have loads of health data to mine. These data can help them train the algorithms to spot patterns in certain data sets of certain cohorts of patients. However, should the wrong or incomplete data sets be used to train the algorithms then the outcomes will be unreliable.


There is a clear need for personalised medicine and one way to rapidly achieve this is through AI. Access to huge data sets and the ability to sift through vast quantities of it rapidly means that healthtech companies are able to develop personalised drug therapies. By looking at data for specific cohorts of people, AI algorithms are able to stratify patient populations and personalise therapies.

Ultimately, I predict that the large pharmaceutical companies will start to recruit the sort of people at these healthtech businesses. They will also start to partner with digital innovation specialists outside of the business that can broaden or deepen the expertise in handling data to find these inventions. If a pharmaceutical company fails to develop a digital technology division or capacity they will be left behind. AI has already changed the way many businesses operate and has successfully proven itself as indispensable in modern business. Now, AI is set to change the pharmaceutical industry through rapidly increasing the speed and range of drug discovery, supporting clinical trials, and driving personalised medicine, and allowing smaller healthtech firms to thrive alongside big pharma.


This article was first published by The European and can be found here.