The Lancet: The higher the education level, the lower the risk of death

This week's noteworthy scientific research

Nature: How does HIV break through the human cell nucleus?

The Lancet: The higher the education level, the lower the risk of death

Nature sub-journal: Scientists have made new discoveries about human aging

AI is prone to making mistakes when faced with the real world, study finds

Nature sub-journal: New materials make AI and computers faster and more efficient

Nature sub-publication: Renewable energy helps produce green hydrogen

How to improve the accuracy of precipitation forecast? Chinese Academy of Sciences: Try "AI + Physics"

Science sub-journal: AI is expected to detect cancer at an earlier stage

Nature: How does HIV break through the human cell nucleus?

A research team from the University of New South Wales in Australia has discovered the mechanism by which human immunodeficiency virus (HIV) overcomes the restriction of the nuclear pore complex (NPC), enters the cell nucleus and establishes infection. The study shows that HIV can enter the cell nucleus through the pores of the nuclear pore complex without the assistance of molecular chaperone proteins. This discovery not only helps humans gain a deeper understanding of HIV biology, but may also provide new insights into other fields such as (gene therapy, etc.).

Original link: https://www.nature.com/articles/s41586-023-06969-7

The Lancet sub-journal: The higher the education level, the lower the risk of death?

A study published in The Lancet Public Health, a subsidiary of The Lancet, revealed that education has a very significant positive impact on life expectancy. For each additional year of education, the risk of death decreases by 2%, six years of primary education can reduce it by 13%, and 18 years of education can even reduce it by 34%. In addition, the study found that the protective effect of education is comparable to risk factors such as healthy diet, non-smoking and moderate drinking. Even in their fifties and seventies, people can still benefit from education. This study, which brings together data from 59 countries, highlights the key role of education in influencing population health worldwide, especially in low- and middle-income countries.

Paper link: https://www.sciencedirect.com/science/article/pii/S2468266723003067?via%3Dihub

Nature sub-journal: New materials make AI and computers faster and more efficient

In future artificial intelligence and data center systems, phase change memory with GST467 nanocomposite materials may become a better choice. The technology is fast, low power, stable, durable, and can be manufactured at temperatures compatible with commercial manufacturing. Researchers at Stanford University said that they not only improved a single indicator, such as durability and speed, but also improved multiple indicators at the same time. This is considered to be a new industry-friendly material built in this field. The research team said this is a key step towards universal memory. The relevant research has been published in the scientific journal Nature Communications.

Figure｜Superlattice phase change memory (PCM) using GST467 nanocomposite material

Original link: https://www.nature.com/articles/s41467-023-42792-4

Nature sub-journal : Scientists have made new discoveries about human aging

A study published in Nature Communications by a team from Nanyang Technological University shows that in the human brain, communication between neurons responsible for memory is disrupted with age. This decline can even begin in middle age. By monitoring the activity of neurons in living mice in real time, the research team found that middle-aged and elderly mice required more training when learning new tasks, and that the connections between neurons in elderly mice weakened, resulting in a weakened ability to retain memory. This discovery provides new clues for developing therapies to maintain cognitive health, while highlighting the importance of early intervention.

Original link: https://www.nature.com/articles/s41467-023-43142-0

Nature sub-publication: Renewable energy helps produce green hydrogen

Researchers from the School of Materials Science and Engineering at the Technion-Israel Institute of Technology published a breakthrough study in Nature Materials - producing green hydrogen from renewable energy. Compared with other processes for producing green hydrogen, this technology uses electrolysis to decompose water into oxygen and hydrogen by using renewable energy such as wind and solar energy. The researchers emphasized that the commercial development of this technology is of great significance for reducing carbon emissions and replacing fossil fuels.

Figure｜Schematic diagram of the processing process

Original link: https://www.nature.com/articles/s41563-023-01767-y

AI is prone to making mistakes when faced with the real world, study finds

A medical study from Northwestern University in the United States found that tissue sample contamination can easily confuse artificial intelligence (AI) models. This problem may cause AI to make mistakes in real environments, especially when the model is trained in an ideal simulation environment. The study explored the impact of tissue sample contamination on machine learning models for the first time. The results showed that four AI models made errors when faced with contaminated tissue, such as in diagnosing vascular damage, estimating gestational age, classifying macroscopic lesions, and detecting prostate cancer. The study calls for strengthening the encoding of biological impurities in AI models to ensure that the models better adapt to the real world, while reminding patients that human experts are still needed to make the final diagnosis.

Original link:

https://www.sciencedirect.com/science/article/pii/S0893395224000024?via%3Dihub

How to improve the accuracy of precipitation forecast? Chinese Academy of Sciences: Try "AI + Physics"

A research team led by Professor Huang Gang from the Institute of Atmospheric Physics, Chinese Academy of Sciences, has improved precipitation forecasting techniques by coupling physical variables with graph neural networks (GNNs). The study utilized the newly developed EarthLab Earth System Science Numerical Simulator facility to improve the precipitation forecasting skills of numerical models. By addressing the difficulties in precipitation forecasting, especially heavy rainfall events, the team constructed a variable coupling graph and introduced physical constraints into the model to improve the accuracy of precipitation forecasts. The results show that this method has significant improvements in precipitation forecasting capabilities in various categories, especially in heavy rainfall events. This result was published in Geophysical Research Letters.

Figure | (a) ɷ-GNN model diagram; (b) details of the encoder and the embedding process of each atmospheric element, n is the dimension of the climate background embedding vector; (c) details of the locality assumption, taking one of the elements as an example; (d) decoder details. Red indicates the corresponding precipitation row vector

Original link:

https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL106676

Science sub-journal: AI is expected to detect cancer at an earlier stage

Using machine learning, researchers from City of Hope and TGen in the United States have developed an innovative small sample blood test technology that is expected to achieve earlier cancer detection. The technology identified half of the 11 types of cancer through the A-Plus algorithm with high accuracy, with only 1 false positive for every 100 tests. The method uses "fragmentomics" technology and only requires one-eighth of the blood volume for whole genome sequencing. In the future, they plan to conduct clinical trials to verify the effectiveness of this technology in earlier cancer stages.

Original link:

https://www.science.org/doi/10.1126/scitranslmed.adi3883