Hello Nature readers,
Today we explore a nuclear-fusion milestone, try to pinpoint the Anthropocene and enjoy some of the best science images of the year.

Nuclear-fusion lab achieves ‘ignition’

Scientists at the world’s largest nuclear-fusion facility have for the first time achieved the phenomenon known as ignition — creating a nuclear reaction that generates more energy than it consumes. The breakthrough at the US National Ignition Facility (NIF) has excited the global fusion-research community. “I’m having a cosmo to celebrate,” says Michael Campbell, former director of a fusion laboratory in New York. The laboratory’s analysis suggests that the reaction released some 3.15 megajoules of energy — roughly 54% more than the energy that went into the reaction, and more than double the previous record of 1.3 megajoules. However, NIF’s 192 lasers consumed 322 megajoules of energy in the process. “It’s a big milestone, but NIF is not a fusion-energy device,” says nuclear engineer Dave Hammer.

“I’m not going to retire in the classic sense”

This month, Anthony Fauci will step down after 54 years at the US National Institutes of Health. Looking back, he sees the progress against HIV as one the most notable achievements in his field. The early 1980s, when the disease was almost a death sentence, “was one of the darkest periods of my or anybody’s professional career in infectious diseases”, says Fauci. But, in just a few years, we went from that bleak time to transformative treatments and highly effective preventive medication.

Protein-folding AI tackles next big challenge

Researchers are repurposing AlphaFold, an artificial-intelligence (AI) system that predicts protein structures, to tackle ever-more-complex challenges in the field. Two years ago, the deep-learning algorithm, developed by DeepMind, swept a competition in which amino-acid sequences were used to predict proteins’ 3D shapes, which are important for determining their function. This year’s competition added harder problems, such as protein–drug interactions and multi-protein complexes. AlphaFold methods aced most tests, but still struggled to predict complexes containing antibodies.

Giant wombat skull hints at distinct lifestyle

A fossilized skull found in a cave in Australia is the 80,000-year-old remains of a giant wombat (Ramsayia magna). A study of the most complete giant wombat skull ever found has revealed that the creature probably had a sizable, fleshy nose and strong muscles for chewing tougher foods, and they did not live in burrows like modern wombats do. The sheep-sized animal probably weighed about 130 kilograms. Scientists still don’t know when or how the giant wombat went extinct.

The place that defines a new Earth epoch

Nine candidate sites are in the running to be the location for the ‘golden spike’ of the Anthropocene, a physical marker for a site that defines a geological epoch. The term Anthropocene informally refers to the current time interval, which is marked by pollution and other signs of human activity — but it isn’t a formal epoch yet. Each spot is being considered for how reliably its mud, ice or sediment preserves markers such as radioactive isotopes from nuclear-bomb tests, ash from fossil-fuel combustion or microplastics. If one is accepted within the next six months, it would end the 12,000-year-old Holocene and officially acknowledge that humans have profoundly changed the planet. Some scientists are opposed to defining an ongoing epoch simply by seeking out the lower boundary of undefined geological layers.

Which single-cell analysis tool is best?

In single-cell research, scientists have to choose from a confusing number of assay methods and nearly 1,400 analysis programs. “Even a small change can lead to substantial differences in the results,” says computational biologist Julio Saez-Rodriguez. The solution: benchmarking. Smart-Seq2, CEL-Seq2 and 10X Chromium identify the most genes per cell, and Quartz-Seq2 particularly excels at grouping cells by biomarker expression. For software packages, scientists should look for methods that perform well with data sets similar to their own — and keep in mind that the most complicated methods aren’t always the best.

The world’s vaccine strategy needs changing

“Despite rallying to produce billions of doses of vaccines in the face of COVID-19, when it comes to developing vaccines to prevent a disease in the first place, the world is still asleep at the wheel,” argues Seth Berkley, chief executive of Gavi, the Vaccine Alliance. The current outbreak of Sudan ebolavirus, which could have been prevented with vaccines, is a perfect example of a global failure to learn from the COVID-19 pandemic, he says. Wealthy countries should take the lead to create an adequate, publicly subsidized vaccine market before the next disaster strikes. “I warned about this problem seven years ago in a column in Nature,” says Berkley. “What will it take to finally catalyse change, so that I’m not writing this again seven years from now?”