When Will AI Beat The Eroom’s Law In The Pharmaceutical Industry?


In 2020, the year of the Covid-19 pandemic, the FDA approved just 53 new drugs. In the same year, the global pharmaceutical industry spent nearly $200 billion in drug research and development costs. This means that the average cost per drug approved in 2020 was nearly $3.8 billion. A study published that year put a more conservative range on the R&D costs of a new drug, noting that the cost of producing a new drug has increased dramatically over the last decade but still ranges between $314 million to $2.8 billion per new drug. This study also found that the median R&D investment required to bring a new drug to the market is nearly $1 billion, while the mean was estimated to be $1.3 billion. It takes on average 10 to 15 years to bring a new drug to market. About half of this time and investment is consumed during the clinical trial phases of the drug development cycle while the remaining half covers preclinical compound discovery, testing, and regulatory processes. Factors that result in such high costs and so many years range from lack of clinical efficacy to lack of commercial interest and poor strategic planning. This makes it difficult to measure the efficiency of the pharma industry, a topic that I have written about and which can be accessed here. The high costs of entering a new drug to market and the time it takes to do so have given birth to a generation of skeptics. These are people who question why the pharmaceutical industry is in its current state despite all of the technological and managerial advancements.

These people are proponents of Eroom’s law, the concept that the cost of developing a new drug has increased exponentially in the last several decades despite improvements in technology. Eroom’s law states that the inflation-adjusted cost of developing a new drug roughly doubles every nine years. This observation is similar to the law of diminishing returns, a concept in economics which suggests that if one input in the production of a commodity is increased while all other inputs are held fixed, a point will eventually be reached at which additions of the input yield progressively smaller, or diminishing, increases in output. The term Eroom’s law was coined by Dr. Jack Scannell and colleagues in 2012 in Nature Reviews Drug Discovery.

Eroom’s law is actually Moore’s law spelled backwards. In case you don’t know what that is, Moore’s law is a concept from the 1960s which observes that the number of transistors in a dense integrated circuit doubles every two years. Moore’s law, which is actually named after Intel co-founder Gordon Moore, is an observation and projection of historical trend.

Dr. Scannell highlights four main causes that have gotten us into this fix, and all four of them are worth discussing. These are: the ‘better then Beatles’ problem, which means a progressively higher bar for improvements over existing therapies; the ‘cautious regulator’ problem, or the progressive lowering of risk tolerance by regulator agencies that make R&D costlier and harder; the ‘throw money at it’ tendency, or the tendency to add other resources to R&D that could lead to project overrun; and the ‘basic research-brute force’ bias, which is the tendency to overestimate the ability of advances in basic research and brute force screening methods.

However, in spite of all these factors, there will come a time when we will finally beat Eroom’s law in the pharmaceutical industry. The solution, among others, is mainly the effective use of AI. And the path towards that is already in the works.

Dr. Scannell and his co-scientists suggest that pharmaceutical companies should appoint a Chief Dead Drug Officer responsible for uncovering the reasons behind a drug failure at each phase of the R&D process, and publish the results in a scientific journal. Pharmaceutical companies today rarely publish the results of failed experiments or clinical trials and it is very unlikely that companies will create a slot for a dead drug officer soon. But this suggestion does highlight that the R&D process needs to change if companies want to beat Eroom’s law. Collaboration and greater sharing of information might be a good place to start. But the only real way of beating Eroom’s law in the pharmaceutical industry is through AI.

Several attempts have been made throughout the last few years to use AI to defeat Eroom’s law and today there are several organizations including Exscientia and Insilico Medicine, that are working towards breaking this law.

Oxford-based Exscientia is a global pharmatech company that uses patient-first artificial intelligence to discover better drugs, faster. Last year, this company announced the first AI-designed molecule for immuno-oncology to enter human clinical trials. In this instance, Exsientia partnered with Evotec to invent and develop an A2a receptor antagonist for adult patients with advanced solid tumors, using Exscientia’s Centaur Chemist drug discovery platform. And this wasn’t even Exscientia’s first rodeo. In fact, the company in 2020 announced that a drug designed by AI powered software entered into a phase 1 clinical trial for the treatment of OCD, or obsessive-compulsive disorder.

Another example is that of Schrodinger, which develops state-of-the-art chemical simulation software for use in pharmaceutical. Schrodinger recently received FDA approval to study its computer-designed therapy for non-Hodgkin lymphomas in an early phase trial. The company’s platform, powered by machine-learning capabilities, sorted through 8.2 billion potential compounds over a 10-month period and identified 78 that were synthesized and filtered through preclinical experiments to select the most promising candidate. Now the company plans to launch its phase 1 clinical study, recruiting patients with relapsed or refractory non-Hodgkin B-cell lymphoma.

Similarly, Recursion Pharmaceuticals in Utah uses AI to find new uses for the drugs owned by other companies. Last year, Roche and Genetech entered into a collaboration with Recursion to explore new areas of cell biology and develop new treatments in areas of neuroscience and an oncology indication. Through the partnership, the companies will use Recursion’s AI-based drug discovery platform to cast a comprehensive net for novel drug targets and expedite the development of small molecule medicines.

At Insilico, we successfully completed a phase 0 clinical study and entered a phase 1 clinical trial with our first-in-class anti-fibrotic drug candidate for a novel target discovered using our AI platform Pharma.AI. The total time from target discovery program initiated to the start of phase 1 took under 30 months, representing a new level in therapeutic asset development speed for the pharmaceutical industry.

Let’s not forget that AI also includes brain-machine interface, deep learning, human-machine interface, machine learning, and other machine simulation of human intelligence. These concepts have been around since decades. Whereas early medical AI systems heavily relied on medical domain experts to train computers by encoding clinical knowledge as logic rules, the technology has now evolved that super computers can do these tasks on their own.

In order to defeat Eroom’s law, data scientists and medical scientists must jointly define achievable use cases where the applications of AI can be used in clinical trials. Such AI technology needs to be tested alongside existing technology it aims to replace or complement. Following this approach, AI may be adopted into the clinical trial ecosystem to rapidly improve the drug discovery and development process in the industry, while also lowering failure rates and the costs. Almost all big pharma corporations are using internal algorithms, partnering with AI companies, or acquiring AI companies to use their technologies and increase their own portfolio and pipeline of drug discovery. Plentiful financing and multiple pharma partnerships showcase the growing interest in using AI tools in the drug R&D process. So we are seeing a lot of movement in this area already, and hopefully in the coming years, when companies begin to combine better investment strategies with advanced AI, they can beat Eroom’s law.