Unlock Editor Digest for free.
Rola Khalaf, editor of the FT, selects her favorite stories in this weekly newsletter.
The artificial intelligence hype cycle depends on flashy announcements that break records. In April, San Francisco-based startup Xaira announced it had raised $1bn in one of the largest biotech launches ever.
Zaira claims that drug development is poised for an AI revolution. It is not alone. Google DeepMind co-founder Damis Hassabis, best known for solving the 50-year-old scientific challenge of predicting the shape of proteins, argues that biology could be “perfect” for AI, because it is fundamentally It is an information processing system. He heads Isomorphic Labs, Alphabet's AI drugs offshoot that has agreed up to $3bn in partnerships with Eli Lilly and Novartis. It aims to halve the drug discovery phase to just two years.
A growing number of AI-derived compounds are in development. The World Health Organization has identified at least 73, although none have yet been approved for use in humans. Some companies are coming close. Ansilico Medicines, which recently filed for a Hong Kong IPO, was the first to get an AI-developed drug into Phase II clinical trials.
But AI is still no substitute for experience that narrows the understanding of a disease. The sector is already struggling. On Xaira's launch day, BenevolentAI announced a major holiday. The London-based company aims to combine human and machine intelligence but has lost 94 per cent of its share price since it was valued at €1.5bn in a December 2021 merger with a special purpose vehicle. has been made public with
Developing innovative new drugs is expensive and inefficient. The pharmaceutical industry has no shortage of funds or motivation when it comes to improving the success rate of drug discovery using AI. According to consultants BCG, around 200 “AI-first” biotechs have raised more than $18bn in the decade to 2023. Both AI usage and success rates vary.
The use of computing in drug design is not new, having originated in the 1970s. Insights are only as good as the data used to train the models. Prediction of the toxicity of drug candidates is hampered by the lack or inconsistency of publicly available information. For example, there is a huge amount of data on lucrative and hotly ongoing research areas like cancer. Less so in relatively neglected areas such as mental health or infectious diseases.
AI is not a magic solution to these problems. Data gaps can be filled through experimentation, but it takes time and deep pockets.
vanessa.houlder@ft.com