A “game-changing” cancer test
to capture the wind. The company says they will be capable of harnessing enough energy to carry 7,000 cars across the Atlantic in 12 days – only five days more than a diesel-powered ship takes. The boat has an auxiliary engine, which could help get it through heavy storms and lulls, but Wallenius Marine says that on a normal voyage, its emissions would be 90% lower than those of a conventional cargo ship. Shipping is estimated to account for 2.8% of all human-made carbon emissions. Per Tunell, the firm’s chief operating officer, described Oceanbird as “something completely different” from ships 200 years ago: “They were sailing with the wind, we are using the wind – which is different.” The company hopes to start production by 2024.
A blood test for cancer that scientists have hailed as a “game-changer” is to be the subject of a major new NHS trial. The Galleri test, developed by California-based company Grail, is one of a new generation of “liquid biopsies” that work by picking up small pieces of tumour DNA in the bloodstream. The company says it can detect over 50 types of cancer and pinpoint where in the body tumours are developing. Starting in the middle of next year, the test will be offered to 165,000 people in England; of those, 140,000 will have no symptoms of cancer, while 25,000 will have been identified by their GPs as having possible signs of the disease, such as lumps or discharges. If the test proves effective at identifying cancers in this cohort, the NHS will move to a larger trial involving a million people, which will take place in 2024 and 2025. Sir Simon Stevens, chief executive of NHS England, said that because the test could lead to earlier detection, particularly of hard-totreat conditions such as ovarian and pancreatic cancer, it had “the potential to save many lives”.
Until now, DeepMind has been best known for its “superhuman performance” in games such as chess and Go, said The Guardian. Now, though, the Google-owned AI company has turned its attention to a rather more serious challenge: predicting how proteins curl up from chains of amino acids into the 3D shapes that enable them to perform specific tasks in the body. This is called the “protein folding problem”. It’s considered a “holy grail” of biology; and DeepMind appears to have cracked it (
Proteins are present in every cell, and are essential to all biological processes. Only by knowing the complex shapes they form can scientists really understand how they work, and develop drugs that can change or alter their behaviour. But determining the shapes is fiendishly difficult: it can take years of lab work to identify the structure of a single protein – and there are more than 200 million in total. For years, computer scientists have been collaborating on this challenge, by trying to create systems to predict the shapes that the amino acids will form.
With billions of possible configurations, it has proved hard; but two years ago, DeepMind started training its AlphaFold algorithm by feeding it data on 170,000 protein sequences and their shapes – and it seems to have worked. In tests, it was able to deduce the shape as accurately as lab-based methods – in a fraction of the time. “I nearly fell off my chair when I saw these results,” said Ewan Birney, of the European Molecular Biology Laboratory. The research must now be peer-reviewed.