The discovery of a potentially groundbreaking medication typically involves extensive laboratory analysis and meticulous examination of data and test results by teams of researchers, a process that can take years.
However, Takeda Pharmaceutical Co., a Japanese company, took a different approach when it acquired an experimental psoriasis drug from a Boston startup for $4 billion in February…
In just six months, Takeda utilized artificial intelligence (AI) to select a promising compound out of thousands of potential molecules.
In the following months, this drug, chosen through AI and machine learning algorithms, will advance to the final stages of clinical trials.
If successful, it could become one of the first therapies developed with the aid of AI. Industry analysts at Jefferies estimate that it has the potential to generate annual sales of up $3.7 billion…
Takeda's strategic move aligns with the broader trend of pharmaceutical companies worldwide embracing AI. They are striking deals with tech-savvy startups and expanding their own teams of data scientists, all in the pursuit of cost reduction and accelerated time to market.
Morgan Stanley predicts that the utilization of AI in early-stage drug development over the next decade could lead to the creation of 50 novel therapies, generating sales exceeding $50 billion.
According to research firm Deep Pharma Intelligence, investments in AI-driven drug discovery companies have tripled in the past four years, reaching $24.6 billion in 2022. Numerous pharmaceutical giants, including Bayer, Roche Holding, and Takeda, are collaborating with Recursion Pharmaceuticals Inc. in Salt Lake City to explore drug discovery using machine learning. AstraZeneca Plc has formed partnerships with BenevolentAI in the UK and Illumina Inc. in San Diego to pursue similar goals.
Alex Devereson, a partner at McKinsey & Co., an advisory firm for drug manufacturers on digital processes and analytics, emphasizes the potential impact of successful AI applications in pharmaceutical research and development. He expects these approaches to become more deeply integrated into pharma R&D processes in the next five years, resulting in significant scale and effectiveness.
While AI can aid in the initial identification of potential drug candidates, scientists still need to conduct extensive traditional research and clinical trials afterward. The Takeda compound, for example, required several additional years of human trials and testing. Moreover, AI has certain limitations, such as its inability to predict complex biological properties like compound efficacy and side effects.
Nonetheless, employing technology to identify promising therapies can reduce the guesswork and time-consuming laboratory experiments that have traditionally been required for such discoveries. The pharmaceutical industry's increased interest in investing in AI and machine learning began in earnest in 2018, when Google's DeepMind unit used its AlphaFold AI program to surpass a biologist's ability to predict protein structures. Protein structure determination is a challenging task in biology, and accurate predictions can help narrow down molecules for interaction and identify disease-targeting medicines.
Bringing a new drug to market typically costs nearly $3 billion, and approximately 90% of experimental medicines fail. Therefore, technologies that expedite the process can have a significant impact on profitability.
AlphaFold, for instance, can determine the 3D structure of a protein in seconds, whereas traditional methods can take months or years. The urgency of the Covid-19 pandemic also accelerated the adoption of AI by the pharmaceutical industry as companies raced to develop treatments and vaccines.
Pfizer, for example, used AI to develop the Covid vaccine Comirnaty, in partnership with BioNTech SE, and to expedite the chemical formulation of the Covid pill Paxlovid, in collaboration with Shenzhen-based AI drug discoverer XtalPi Inc. Both treatments received approval from the US Food and Drug Administration in less than two years, a significantly faster timeline compared to the usual 10 years for most drugs. Regulatory agencies also prioritized the approval of Covid-related treatments.
Takeda's acquisition of an experimental drug from Nimbus Therapeutics LLC, based in Boston, could lead to one of the few oral treatments for psoriasis, a condition affecting 125 million people globally.
Additionally, the drug shows potential for treating other conditions such as Crohn's disease. Named TAK-279 for now, the compound has already successfully completed the first two stages of human trials. Instead of having scientists test an overwhelming number of molecules, the AI-based approach suggested testing only 10 compounds in the lab and learning from the results to refine predictions for the next batch of candidates.
Currently, Takeda's more than 500 quantitative scientists and tech experts across its R&D centers in Boston, San Diego, and Shonan, Japan, devote their time to data analysis in order to discover, develop, and manufacture breakthrough medicines. The company collaborates with the Massachusetts Institute of Technology and multiple AI startups. Anne Heatherington, head of Takeda's data science institute, stresses the importance of technology that empowers employees, reduces manual work, eliminates friction in the system, and allows for greater scientific insight and discovery.
Other major pharmaceutical companies, like Pfizer, are also embracing AI. Pfizer expects its partnership with DeepMind's AlphaFold to facilitate the design and validation of highly effective therapeutic targets that were previously unknown. The company's chief digital and technology officer, Lidia Fonseca, highlights the use of powerful supercomputing capabilities, AI, and machine learning models, which reduced computational times by 80% to 90% and expedited the development of Paxlovid.
Around the world, several potential drugs discovered with the aid of AI are already undergoing human trials. For example, Recursion Pharmaceuticals Inc. is working on five candidates for rare diseases and oncology, while Exscientia is developing three drugs for conditions like cancer and inflammation. Insilico Medicine, based in Hong Kong, has a candidate in midstage human trials for treating the most common form of pulmonary fibrosis.
GSK Plc, headquartered in the UK, has over 160 experts dedicated to AI and machine learning, supporting their R&D and manufacturing efforts. The company leverages data to build and feed its own machine learning models, enabling all scientists to benefit from past data produced by the company. China is also investing in AI to enhance the global competitiveness of its drugmakers. XTalpi, for instance, receives partial funding from Chinese tech giant Tencent Holdings Ltd., and Baidu Inc. CEO Robin Li established BioMap, an AI-driven drug discovery company.
Although AI excels in synthesizing data from various sources, challenges arise when dealing with complex systems, according to Kim Branson, head of AI at GSK. To ensure safety, laboratory experiments remain crucial.
It is important to note that the data used to create algorithms may contain biases, which can be reflected in the clinical recommendations generated. Researchers at Stanford University highlighted this concern in a 2018 study published in the New England Journal of Medicine. The study also revealed that algorithm results can be influenced by their developers.
Despite these challenges, investments in AI for drug discovery continue to surge. Russ Altman, a bioengineering professor at Stanford with years of experience conducting due diligence for venture capitalists in the biotech sector, has witnessed a significant increase in requests to evaluate potential AI drug discovery companies over the past five years. The trend has seen rapid growth, creating an active and expanding field of interest for investors.
On March 1, 2022, a tiny biotech firm won a patent on a new treatment that has the potential to “Cut & Paste” disease from your body.
The Wall Street Journal reports this company is “transforming medicine.”