Unlock Editor's Digest for free
FT editor Roula Khalaf has chosen her favorite stories in this weekly newsletter.
The artificial intelligence hype cycle relies on flashy record-breaking announcements. In April, San Francisco-based startup Xaira did just that, announcing it had raised $1 billion in one of the largest launches in biotech history.
Zaira argues that drug development is poised for an AI revolution. It's not alone. Demis Hassabis, co-founder of Google DeepMind and famous for solving his 50-year-old scientific challenge of predicting the shape of proteins, said that since biology is fundamentally an information processing system, AI is It claims it could be “perfect.” He heads Isomorphic Labs, Alphabet's AI pharmaceuticals spin-off, which has agreed to partnerships worth up to $3 billion with Eli Lilly and Novartis. He aims to halve the drug discovery stage to just two years.
The number of AI-derived compounds in development is increasing. The World Health Organization has identified at least 73 types, but none have yet been approved for human use. Some companies are getting close to that. Insilico Medicine, which recently filed for an IPO in Hong Kong, is the first company to bring an AI-designed drug into Phase II clinical trials.
However, AI still cannot replace the experiments that support our understanding of disease. The sector is already experiencing problems. On the day of Xaira's launch, BenevolentAI announced major layoffs. The London-based company, which aimed to combine human and machine intelligence, has seen its shares lose value since going public in December 2021 at a valuation of €1.5 billion through a merger with a special acquisition vehicle. I lost 94%.
Developing innovative new drugs is expensive and inefficient. The pharmaceutical industry does not lack funding or motivation when it comes to using AI to improve drug discovery success rates. About 200 “AI-first” biotech companies have secured more than $18 billion over the decade to 2023, according to consultancy BCG. AI usage and success rates both vary.
The use of computing in drug design is by no means new, with origins dating back to the 1970s. Insights are only as good as the data used to train your model. Prediction of toxicity for drug candidates is hampered by the lack or relevance of publicly available information. For example, there is a wealth of data on lucrative and intensely researched areas of research, such as cancer. Less is known about relatively neglected areas such as mental health and infectious diseases.
AI is not a magic solution to these problems. Data gaps can be filled through experimentation, but it takes time and plenty of money.
vanessa.holder@ft.com