24-25 - Artificial intelligence: a smart bet for the pharma industry

For the second year in a row, the JP Morgan Healthcare Conference was held as a fully virtual conference in 2022. What the event lacked this year in elbow-rubbing and after-hour receptions, it made up for in a steady pouring of R&D deals, scientific collaborations and digitally-laden growth strategies that show the steps and progress many organisations are taking to transform how they build healthcare products, deliver health services and engage with patients and consumers.

Transformation requires real reasons for change and there are many reasons across the healthcare ecosystem: physician burnout, rising healthcare costs, changing patient and consumer preferences, uncured disease and, as we enter into the third year of the pandemic, the ever-present reminder of infectious diseases and the stress they place on healthcare systems and society.

Healthcare organisations are turning to artificial intelligence (AI) to buck some of these trends. A recent Gartner prediction report states that by 2023, 20% of all patient interactions will involve some form of AI enablement, up from less than 4% today. Last year, venture capital deals in the United States reached an all-time high of nearly $330bn, with the top bets in technology, biotech, healthcare and fintech for a record 17,054 deals.
NVIDIA works with leaders across the healthcare industry around the world to implement AI through its accelerated computing platform, and the company sees four key areas where AI is likely to gain traction and drive forward the transformation in the healthcare industry.

The pharmaceutical industry has grown accustomed to investing billions of dollars to bring drugs to market, only to watch 90% of them fail even before they make it to clinical trials. Due to recent AI powered scientific breakthroughs, R&D leaders are turning to AI to identify the right targets and increase the likelihood of building successful molecules.

The simultaneous breakthroughs of AlphaFold and RoseTTAFold are creating a thousand-fold explosion of known protein structures. Generative AI can conjure thousands more potential chemical compounds outside known databases and has increased the opportunity to discover drugs by a million times.

Transformer-based neural networks can master the ‘language’ of biochemistry, using Natural Language Processing (NLP) to learn from huge data sets of protein sequences, effectively learning the syntax and grammar of biology in a self-supervised manner, without the need for labour intensive labelling.

Similarly, graph-transformer models can be trained on large datasets that can be represented as graphs, such as compounds, and learn to execute a variety of downstream tasks such as property prediction or molecule generation.
Biopharmaceutical company AstraZeneca and NVIDIA collaborated on a transformer-based AI generative model for chemical structures, being run on Cambridge-1, the UK’s most powerful AI supercomputer.

At the JP Morgan Healthcare conference, pharma giant Merck announced a collaboration with Biotech company Absci that could lead to a windfall of $600m in payments for Absci for using AI to expand the potential of proteins and development of new therapeutics.

Amgen and generative biology company, Generate Biomedicines also revealed a new $1.9bn deal that will use AI and machine learning to discover new drugs.
AI as a discovery tool is quickly becoming a viable path to identifying new targets, building new molecules, and ultimately speeding up the development of effective therapies for patients. It’s only natural that early-stage companies founded on these advanced computational methods are starting to reap the benefits of those early bets.
Collaborations between upstart biotech and established pharma companies are by no means new, but it will be interesting to watch how many organisations take a partner-first approach and how many start to ramp investments to accelerate their own AI programmes.

Drug discovery is a data-intensive process where researchers labour on computationally dense calculations to simulate how molecules interact to identify the right therapeutics.

Physics-based in silico approaches to drug discovery such as molecular dynamics can enable fast, accurate and cost-effective binding and other molecular property predictions. But these accurate simulations, like quantum mechanics, are oftentimes too computationally intensive to perform given reasonable time, cost and computing resources.

But deep learning approaches are now able to learn the fundamentals of physics, allowing scientists to project highly accurate simulations based on the laws of chemistry for 1,000 times less computational cost.

A new breed of tools is emerging both from academia and industry that aims to bring AI-powered simulation to the forefront. One such organisation is Entos, that is revolutionising small molecule design through its Orbnet technology, a deep learning architecture that provides quantum level simulations at a significantly reduced computational expense.

The medical device industry has an opportunity enabled by AI to miniaturise device size and reduce cost, to automate and increase accessibility and continuously deliver innovation over the life of the product.

Leaders in this space are looking to increase the pace of innovation while adopting new technology and simultaneously focusing on lowering product development costs.

This is creating a new business model to enable medical device companies to evolve from hardware solutions into software-as-a-service solutions, which can be upgraded remotely which will keep devices in state-of-the-art condition years after deployment.

NVIDIA Clara Holoscan is the computational infrastructure developers can build on to stream data from medical devices and keep them constantly updated to build next-generation healthcare and medical applications.

AI as a technological approach toward value creation in healthcare was perhaps more prominent at this JP Morgan than ever before. Sequencing companies had a banner year as they’ve proven to be an instrumental technology in the fight against the pandemic. These same companies consistently cited the investment in algorithmic advancement as key strategic directives for R&D.

Companies like UK-based Oxford Nanopore Technologies are building rapid sequencing tools while generating large swaths of genomics data. Investments both in chemistry and in algorithmic advancement have led to notable improvements in read accuracy over the past few years. These instruments, and others like it, are increasingly software defined and AI enabled, continuously increasing the value they provide their users over the life of the instrument.

The healthcare industry faces data governance challenges more acutely than any other industry. A multi-hospital initiative sparked by the COVID-19 crisis illustrated that any industry can develop predictive AI models while also setting a standard for accuracy and generalisability.

The technique, known as federated learning, used an algorithm to analyse chest x-rays and electronic health data from hospital patients with COVID-19 symptoms. The patient data was fully anonymised and an algorithm was sent to each hospital so no data left the location.

The EXAM model, published in Nature Medicine, is among the largest, most diverse clinical federated studies known to date. NVIDIA worked with Mass General Hospital, Cambridge University and several other institutions to make the model publicly available for all to collaborate with using NVIDIA FLARE.

NVIDIA sees this as the future for healthcare as well as other industries to collaborate, build cutting-edge algorithms whole respective data governance challenges.

These modern, AI-fuelled approaches to privacy preserving methods, software-defined medical devices and to drug discovery will change the way scientists and pharmaceutical companies reimagine what is possible. AI and simulation are proving to be essential tools for modern research and computing and pharma companies that fuse the application of data strategy, simulation, and AI will deliver much needed medicines faster.

Artificial intelligence: a smart bet for the pharma industry

Million X drug discovery generated by AI

New Models

AI guided simulation

AI creates SaaS medical devices

Building powerful AI models across healthcare institutions and regions