Nature Portfolio

Nominations are now being accepted for the 2025 John Maddox Prize. The award recognises the work of any individual who promotes science and evidence on a matter of public interest in the face of hostility. https://lnkd.in/eVXX9hD2

A paper in Nature describes the discovery of a new preclinical compound with strong antifungal activity against multidrug-resistant pathogens. The drug, named mandimycin, is a member of a known family of bacterial products with antifungal properties, the polyene macrolides. Read the article: https://lnkd.in/eU-Mb2PC

We're #hiring a new Associate Editor or Senior Editor, Nature Sensors in Jersey City, New Jersey. Apply today or share this post with your network.

People with social disadvantage, during both early and later life, have an increased risk of 66 age-related diseases mediated by 14 age-related proteins, according to a paper in Nature Medicine. Read the article: https://lnkd.in/eaS9FVqX

Felice Frankel writes in an essay for Nature what she has learned while experimenting with AI image generators and what generative AI could mean for the future of science photography.

A paper in Nature Chemical Biology presents a system that uses E. coli bacteria to create biodegradable plastics. The system may help in the production of plastics with desirable thermal and mechanical properties, using renewable resources. Read the article: https://lnkd.in/eprXX8pt

How quickly can we see health returns from reducing air pollution? Over the past three decades, China’s exponential economic and urban growth has brought severe air pollution. But in 2013, China took bold action with the Air Pollution Prevention and Control Action Plan (APPCAP). This is a nationwide, multi-sectoral strategy to tackling air pollution by integrating emissions control, clean energy promotion, tax policies and air quality monitoring. Within just 5 years, analyses of hospital records during and after APPCAP implementation shows a substantial drop in hospital admissions for a wide range of diseases- from depression to lower respiratory infections to cardiorenal disease, all linked with improved air quality. We can learn lessons here- when clean air policies are stringent and multisectoral, they can deliver rapid and wide-ranging health returns. The article is here: https://lnkd.in/e6nTmmiG #airpollution #airquality #PM2.5 #blackcarbon #greenhousegases #carbonemissions #health #noncommunicablediseases

A study in Nature Microbiology finds that the FDA-approved antiviral baloxavir improves disease outcomes in mice infected with avian influenza A (H5N1) virus, compared with the current standard treatment of oseltamivir. Read the article: https://lnkd.in/eYEDb2qQ

In its eighth year, our annual photo competition #scientistsatwork is open for submissions. 'Scientists at work' features amateur photography from researchers featuring scientists at work from a wide range of backgrounds and places. Winners will receive £500 (or equivalent in a different currency) as well as a print and online personal subscription to Nature. They’ll also have their photo featured alongside an interview in an upcoming issue of Nature. Deadline is March 28th and to take part all you will need do is email your best photograph to photocompetition@nature.com. Nature Portfolio Nature MagazineSpringer Nature https://lnkd.in/eBZ-dvYY

🚨 New Paper Alert! 🚨 I am thrilled to share our latest work, published in Nature Communications Medicine! 🎉 🚨 A key question led us to this research. We have rigorous software testing frameworks to catch bugs in traditional programs. But do we have a similar systematic testing framework for machine learning models—especially in healthcare, where decisions can mean—life and death?? 🤔 Don't worry. We introduced a new way to test machine learning models and found serious deficiencies in existing models. ✌ 🔎 Key Findings: 📌 Current model testing is insufficient. Traditional ML testing relies on small test sets that do not represent real-world emergency cases. We propose synthetic test cases to improve evaluation. 📌 66% of severe injuries go unrecognized by tested ML models. Our study found that models failed to assign high mortality risk scores in critical conditions, highlighting dangerous blind spots in clinical AI. 📌 Gradient-based testing reveals major gaps in ML responsiveness We developed a gradient ascent method to systematically probe machine learning models for their ability to recognize deteriorating health conditions. 📌 ML models fail to recognize worsening patient conditions. Using gradient-based synthetic test cases, we found that most ML models do not adjust mortality risk scores adequately even when patient vitals deteriorate significantly 📉 Models struggle to detect critical changes in vital signs. For example, we tested how much systolic blood pressure, respiratory rate, and body temperature must worsen before models flagged risk. Shockingly, many models remained unresponsive even in life-threatening situations. Paper link: https://lnkd.in/e6KnEmSU