By Karen Taylor, Director, Deloitte Centre for Health Solutions

While drug discovery has led to many lifesaving and life-enhancing clinical treatments, it is also a long, expensive and often unsuccessful process, with many areas of unmet need. In 2019 we published our report Intelligent drug discovery: Powered by AI, which explored the rise of AI drug discovery disrupter companies and the challenges and opportunities in using AI technologies to help find new, more precise, targeted treatments more quickly.¹ This was the second in a series of reports exploring the role of AI in transforming the pharmaceutical (pharma) value chain.² Fast forward a few years to February 2023 and the US Food and Drug Administration (FDA) have granted orphan drug designation to a drug discovered by one of the companies we featured as a case study in our 2019 report, that uses a generative AI platform.³ This development and the hype surrounding generative AI provided me with a great opportunity to use this week’s blog to revisit the role of generative AI in drug discovery.

Why accelerating drug discovery is crucial

Over the past 13 years we have published an annual report exploring the performance of the biopharmaceutical (biopharma) industry in generating returns from investment in innovative new medicines. Our latest report, Seize the digital momentum: Measuring the return from pharmaceutical innovation 2022 examines the current state of biopharma research and development (R&D) across the top 20 biopharma companies by R&D spend and found that after the notable rise in the internal rate of return (IRR) in 2021, 2022 saw a return to declining rates of return (to 1.2 per cent).⁴ This decrease reflects the ongoing realities of the challenges facing the industry, specifically, increasing costs with declining returns.

While the pace of innovation in science and technology has accelerated exponentially, the discovery and development of new drugs remains long and expensive, with many failures along the way. The average time to bring a molecule to launch is 10-12 years (with 5-6 years accounted for by the preclinical, drug discovery process). In 2022, the average cost of R&D for the top 20 biopharma companies was $2.284 billion per drug – more than double the $1.188 billion calculated in 2010. On top of this, the average forecast peak sales per late-stage asset declined to $389 million, under half the 2010 value of $816 million. Moreover, cycle times, especially for oncology assets, have increased. Consequently, finding new ways of improving the efficiency and cost-effectiveness of R&D remains one of pharma’s most pressing challenges. As we wrote in 2019, a crucial way of achieving this is by improving the accuracy, predictability and speed of drug discovery, which accounts for around a third of the costs.

The rise of AI drug discovery disruptors

In 2019 we highlighted the emergence of a growing number of AI-enabled solutions with the potential to accelerate drug discovery. Specifically, increasing volumes of structured and unstructured scientific data meant AI could be used to help improve our understanding of structures and specificity to the target molecules. This potential was illustrated by a number of case studies that addressed the five main challenges where companies using AI for drug discovery were focusing their efforts:

• finding new disease-associated targets
• screening small molecule libraries to identify new drug candidates
• de novo design of new drug candidates
• drug optimisation and repurposing
• preclinical testing.

Several of these case studies highlighted the use of generative AI. For example, Insilico Medicine began working with generative AI in 2018, aiming to generate novel chemical compounds that could be developed into new medicines to treat diseases. At the time of our report, it had developed a new platform, Generative Tensorial Reinforcement Learning (GENTRL), which combined two distinctive deep learning (DL) models. Its AI Imagination for drug design, ‘imagines’ molecules with specific properties by using a generator to produce images with selected characteristics and competing this with a discriminator to test if the output is true or false. Once a target is identified scientists use the DL algorithms to design molecules with desired physical and chemical properties. In 2019, the GENTRL platform had generated new drug hits against fibrosis in 21 days and validated them selecting one lead candidate in another 25 days. The process from beginning of the design to process was 15 times less than traditional biopharma timings.

What happened next?

In 2020, scientists at Insilico Medicine launched the Chemistry42 platform (GENTRL was the first incarnation of Chemistry42). It is an automated machine learning (ML) platform connecting generative AI algorithms with medicinal and computational chemistry methodologies to generate novel molecular structures with optimised properties. The generative platform, which encompasses a wide range of networks and models, has been leveraged by over 20 pharma companies and more than 15 external and 30 internal programs. A major advantage of the system is its customisable reward function. As the molecular structures are generated, they are dynamically assessed using the reward function and 3D physics-based modules. Each module scores the generated molecules and together with generative algorithms, optimises those molecules that are most likely to succeed in terms of potency, metabolic stability, synthetic accessibility and more. The novel molecules are further ranked based on their ADME (absorption, distribution, metabolism and excretion) and selectivity profiles’.⁵

In February 2023, Insilico Medicine announced that the US FDA has granted Orphan Drug Designation to ‘INS018_055’, a potentially first-in-class small molecule inhibitor for the treatment of Idiopathic Pulmonary Fibrosis, a chronic lung disease that causes progressive and irreversible decline in lung function and represents a significant unmet medical need worldwide. The FDA's Orphan Drug Designation programme supports the development and evaluation of drugs that address rare diseases which affect fewer than 200,000 people in the United States. Receiving orphan drug designation from the FDA facilitates the subsequent development and commercialisation that comes with the designation, including eligibility for federal grants, tax credits for qualified clinical trials, prescription drug user fee exemptions, and a seven-year marketing exclusivity period upon FDA approval.

The future of drug discovery: delivering more precise targeted treatments

As the above example, and the other case studies in our 2019 report show, if adopted at the drug discovery stage, AI solutions have the potential to kick-start the productivity of the entire R&D process. Indeed, Insilico Medicine’s use of generative AI for target discovery and drug design meant the total time from target discovery to Phase 1 took under 30 months, a new level of speed in therapeutic asset development for the pharmaceutical industry.⁶ Using generative AI can optimise the properties of the molecules faster with fewer iterative cycles in the design phase.

AI and other innovative technologies using data from multiple sources are enabling more precise, targeted treatments and will help shift the health ecosystem towards a future where medicine is more precise, personalised, and predictive. This in turn will enable more efficient and effective models of care. Over the next seven years these shifts will have a significant impact on treatments opening up a new future for the health industry, as a higher level of knowledge on disease mechanisms increases the number of treatments available and, in many cases, cure diseases that have not previously had effective treatments.

Karen Taylor - Director, UK Centre for Health Solutions

Karen is the Research Director of the Centre for Health Solutions. She supports the Healthcare and Life Sciences practice by driving independent and objective business research and analysis into key industry challenges and associated solutions; generating evidence based insights and points of view on issues from pharmaceuticals and technology innovation to healthcare management and reform.

Email | LinkedIn

__________________________________________________________________________________________________________

¹ Using artificial intelligence in biopharma | Deloitte Insights

² Intelligent Biopharma Series | Deloitte UK

³ (24) FDA Grants Orphan Drug Status For Drug Discovered Using Generative AI | LinkedIn

⁴ Seize the digital momentum: Measuring the return from pharmaceutical innovation 2022

⁵ (24) How Insilico Medicine Leverages Generative AI To Discover New Drugs | LinkedIn

⁶ (24) FDA Grants Orphan Drug Status For Drug Discovered Using Generative AI | LinkedIn