Skip to content Skip to sidebar Skip to footer

Source: The Guardian as of 02-04-2020

AI companies and scientists are cooperating, but they desperately need access to pharmaceutical companies’ data

  • Prof Ara Darzi is a surgeon and director of the Institute of Global Health Innovation at Imperial College London

The past few weeks have revealed the worst and the best in human responses to the coronavirus crisis – from the supermarket hoarders clearing the shelves to the neighbourhood groups organising help for elderly and vulnerable people. When it comes to the pharmaceutical companies, how should we judge their response? They, after all, hold the key to ending the pandemic. Yet in one vital respect their behaviour has more in common with the supermarket hoarders than the neighbourhood groups. Our exit strategy from the global lockdown depends on the development of an effective vaccine, as is well-known. A huge effort is under way to find such a vaccine, but we cannot afford to wait the 18 months it might take.

In the meantime, as the death toll increases, doctors are desperate for treatments that would lessen the impact of the virus, by shortening the infection, reducing its severity and in that way saving lives. There is now a global hunt for a coronavirus drug. But it is a fight against time. The focus is therefore on existing treatments already proved to be safe for other diseases which will need less testing and be easier and quicker to manufacture in quantity. Scores of trials are under way around the world. The World Health Organization has identified four of the most promising therapies – including an HIV combination treatment, an anti-malarial and a drug developed but never used against Ebola – for testing in a global trial launched last month. But we cannot pause the search while waiting for the results. The need for new effective agents is too great.

The best way to identify candidate drugs is to use artificial intelligence (AI) to crunch huge quantities of data to find the ones that might work. Major AI companies are putting their immense computing power at the service of scientists engaged in this hunt. But they are being hampered: because some pharmaceutical companies are failing to share all of the data on potential candidate treatments that they hold. Like toilet roll profiteers, they are keeping it stashed in their digital attics and cellars where others cannot get at it, on the grounds that it is commercially confidential.

It was the open sharing of data around the world that allowed scientists to map the genome of the SARS-CoV-2 virus at unprecedented speed, working across institutional, commercial and international boundaries in a unique collective effort against a common global enemy. We now urgently need all pharmaceutical companies to set aside their individual commercial ambitions and join a similar collective effort to identify, test, develop and manufacture treatments to curb the disease. There is a precedent. Last June, 10 of the world’s largest pharmaceutical companies – including Johnson & Johnson, AstraZeneca and GlaxoSmithKline – announced they would pool data for an AI-based search for new antibiotics, which are urgently needed as antibiotic-resistant bacteria have proliferated across the world, threatening the growth of untreatable disease.

That historic agreement was made possible by the development of a secure, blockchain-based system that allows an algorithm to search rival companies’ data with full traceability – but without revealing commercial secrets to competitors. The advantage of using blockchain is that companies can trust the code rather than their partners. AI researchers at the Massachusetts Institute of Technology’s J-Clinic, who trained a neural network to predict which molecules will have antibiotic properties, announced in February that they had found a new compound which works against 35 different types of bacteria. They named it halicin, after the AI system in 2001: A Space Odyssey.

Show CommentsClose Comments

Leave a comment

News ORS © 2020. All Rights Reserved.