COVID-19 Calamity Has Given AI an Opportunity to Shine

1st October 2021

Drug development is an expensive and lengthy process

It takes an average of 10 years for a new medicine to complete the journey from initial discovery to the marketplace, with the clinical trial itself taking around six to seven years. Additionally, the financial stakes are high, with the R&D of any successful drug development costing an average of $2.6 billion. Yet this overall process has seen minimal evolution over time, keeping it unwieldy, inefficient and costly.

Before COVID, the adoption of new technology in the industry was slow; however, the pandemic has prompted increased and revolutionary use of sophisticated technologies in drug development, such as AI, which is now being used in unprecedented ways in response to the crisis. Examples of this include the repurposing of a rheumatology drug for COVID, the identification of a protein structure to design vaccine candidates and the use of AI in clinical trials to reduce failure rates.

Advances in AI-based computational modeling have helped develop novel RNA-based treatments such as mRNA vaccines. Tools have made it possible to design a vaccine containing both B- and T-cell epitopes, while natural language processing (NLP), an AI capability, has been used to predict protein interaction and model molecular reactions in carbohydrate chemistry, not only helping to speed up vaccine development, but also making it possible to use synthetic chemistry to design new drugs. AI has also been used for vaccine research, where processes like virtual screening (VS) are replacing inefficient lab-based, high-throughput screening methods, and it has played a critical role in connecting complex data sets to generate real-world evidence (RWE) at an unprecedented pace.

New Drug Development :In this evolving domain, the computational capabilities of AI have been used for developing drugs and applying core biochemistry principles to test the molecular properties, interactions and behaviors of materials, systems and processes for specific functions. For example, SRI International has partnered with Iktos, a French AI firm, to design and test new potent molecules, utilizing a fully automated system based on synthetic chemistry​.

Drug combinations​: Drug combinations often result in increased therapeutic efficacy and reduced toxicity; however, the effort needed to find a suitable pairing is huge, with infinite possible combinations. Using AI capabilities, potential combinations for COVID have been identified, such as sirolimus with dactinomycin, mercaptopurine with melatonin and toremifene with melatonin. These combinations are based on theoretical analysis and now need to be clinically tested to collect RWE.

Vaccine research: In addition to VS, a method where a specific molecule is computationally targeted to inhibit cell growth, AI is also aiding reverse vaccinology, a framework that accelerates the development process. Meanwhile, machine learning is assisting with predictions for compound properties, activities and reactions and ligand-protein interactions to advance the vaccine discovery process. The use of AI-based automatic extraction features to support models is resulting in better accuracy and more reliable results in much less time compared to traditional methods. It also helps in identifying more druggable molecules, thus reducing the chance of failure as well as mitigating the novelty of newer viral strains by leveraging the transfer of knowledge gained from previous tasks. Lastly, AI has also been used in drug discovery design solutions like Cyclica’s PolypharmDB platform, which has helped uncover off-target applications of 30 existing drugs against the viral protein ACE2, to which the SARS-CoV-2 virus binds.

RWE gathering and analysis: In this evolving domain, the computational capabilities of AI have been used for developing drugs and applying core biochemistry principles to test the molecular properties, interactions and behaviors of materials, systems and processes for specific functions. For example, SRI International has partnered with Iktos, a French AI firm, to design and test new potent molecules, utilizing a fully automated system based on synthetic chemistry.

Repurposing drugs:Drug combinations often result in increased therapeutic efficacy and reduced toxicity; however, the effort needed to find a suitable pairing is huge, with infinite possible combinations. Using AI capabilities, potential combinations for COVID have been identified, such as sirolimus with dactinomycin, mercaptopurine with melatonin and toremifene with melatonin. These combinations are based on theoretical analysis and now need to be clinically tested to collect RWE.

How can AI demonstrate value?

With the recent emphasis on value-based procurements from payers, there has been increased pressure on drug manufacturers to demonstrate value in their products, which can entail many dimensions. One of the most important and impactful metrics is RWE, but this can be difficult to quantify through traditional means because of its dynamic nature and lack of standardized measures and terminology. AI plays an important role by crunching numbers and plotting several different models that can not only demonstrate the value of a product in the immediate future, but also, with the help of its predictive capabilities, identify potential future trends and risks. Furthermore, NLP capabilities that can analyze complex real-world data from speech, text and other media and use those data to plot models for practical purposes have enabled the introduction of new drugs in record time, particularly in COVID.

AI is still evolving. It is a beast with many components, but not all components are needed for everyone. Based on their specific requirements, companies can leverage the components that are most relevant to them; for example, R&D companies can leverage the discovery capabilities of AI for the analysis of virus protein structures, while big global drug manufacturers can leverage components around production, supply chains and inventory management to streamline the practical issues that ensure continued widespread access to COVID drugs. According to McKinsey Global Institute’s estimates, AI and machine learning in the pharmaceutical industry could generate nearly $100 billion annually across the US healthcare system alone, which demonstrates the magnitude of their impact.

Being a niche tech, it is not easy to build AI systems from scratch, especially for pharma companies that do not have a core focus on tech development; however, there are three main ways these companies can achieve this:

Collaboration: With tech groups that have been scouted out for their mutual interests; for example, Alnylam partnered with Paradigm4 for drug development, and in 2019, Novartis and Microsoft signed a multi-year contract.

Acquisition: Some promising fledgling AI start-ups or small companies and establish in-house capabilities that can then be controlled and molded to suit the company’s needs; for example, in 2018, Roche acquired Flatiron Health for $1.9 billion, while Charles River Laboratories acquired Distributed Bio for $83 million.

Development: In-house capabilities that are best aligned with the company needs; for example, GSK was an early pioneer, setting up an in-house AI unit in 2017, with other companies following soon after.

Like evolution, technology automation is unstoppable

AI has proven to be a powerful weapon in this race against time, with new COVID mutations emerging and causing subsequent waves. To meet these new challenges, AI will continue to evolve, and new players will use these evolutions in different ways. According to a report by Research and Markets, global AI in pharma is expected to grow from $0.91 billion in 2020 to $1.27 billion in 2021, and is predicted to reach $5.94 billion in 2025 at a CAGR of 47%. It is therefore imperative to keep track of what happens in the space and keep a close eye on the developments of new emerging players across the globe. This will not only help in anticipating what competitors are working on, but also in finding potential players to collaborate with.