How artificial intelligence learns from complex networks

Multi-layered, so-called deep neural networks are highly complex constructs that are inspired by the structure of the human brain. However, one shortcoming remains: the inner workings and decisions of these models often defy explanation.
Image Credit: Brian Penny / AI Generated

Deep neural networks have achieved remarkable results across science and technology, but it remains largely unclear what makes them work so well. A new study sheds light on the inner workings of deep learning models that learn from relational datasets, such as those found in biological and social networks.

Graph Neural Networks (GNNs) are artificial neural networks designed to represent entities—such as individuals, molecules, or cities—and the interactions between them. These networks have practical applications in various domains; for example, they predict traffic flows in Google Maps and accelerate the discovery of new antibiotics within computational drug discovery pipelines.

GNNs are notably utilized by AlphaFold, an acclaimed AI system that addresses the complex issue of protein folding in biology. Despite these achievements, the foundational principles driving their success are poorly understood.

A recent study sheds light on how these AI algorithms extract knowledge from complex networks and identifies ways to enhance their performance in various applications.

Scientists Have Created Organic Films to Charge Cardiac Pacemakers

The resulting films have high biocompatibility.
Photo Credit: Andrei Ushakov

UrFU scientists, together with colleagues from the University of Aveiro (Portugal), have succeeded in obtaining biocompatible crystalline films. They have high piezoelectric properties - they generate an electric current under mechanical or thermal stress. This property will be useful in the design of elements for invasive medical devices, such as pacemakers. Detailed information about the films obtained and the new method of their synthesis has been published by the scientists in ACS Biomaterials Science & Engineering. 

"We have succeeded in obtaining films from diphenylalanine that have high piezoelectric properties comparable to their inorganic counterparts. Under mechanical or thermal stress, these films generate electricity. The use of such films will be particularly useful for making invasive cardiac pacemakers - devices that reside inside the human body. When the heart moves or beats, these films generate electricity, which is stored in the pacemaker's batteries. Energy storage devices based on such materials could solve the problem of replacing depleted batteries and reduce the number of surgical procedures," explains Denis Alikin, Head of the Laboratory of Functional Nanomaterials and Nanodevices at the UrFU Research Institute of Physics and Applied Mathematics.

New dressing robot can ‘mimic’ the actions of care-workers

The world's first bimanual dressing robot system mimics how caregivers assist humans in dressing.
Photo Credit: Courtesy of University of York

Scientists have developed a new robot that can ‘mimic’ the two-handed movements of care-workers as they dress an individual.

Until now, assistive dressing robots, designed to help an elderly person or a person with a disability get dressed, have been created in the laboratory as a one-armed machine, but research has shown that this can be uncomfortable for the person in care or impractical. 

To tackle this problem, Dr Jihong Zhu, a robotics researcher at the University of York’s Institute for Safe Autonomy, proposed a two-armed assistive dressing scheme, which has not been attempted in previous research, but inspired by caregivers who have demonstrated that specific actions are required to reduce discomfort and distress to the individual in their care.

It is thought that this technology could be significant in the social care system to allow care-workers to spend less time on practical tasks and more time on the health and mental well-being of individuals. 

DNA Aptamer Drug Sensors Can Instantly Detect Cocaine, Heroin and Fentanyl – Even When Combined with Other Drugs

Photo Credit: Nastya Dulhiier

Researchers from North Carolina State University have developed a new generation of high-performance DNA aptamers and highly accurate drug sensors for cocaine and other opioids. The sensors are drug specific and can detect trace amounts of fentanyl, heroin, and cocaine – even when these drugs are mixed with other drugs or with cutting agents and adulterants such as caffeine, sugar, or procaine. The sensors could have far-reaching benefits for health care workers and law enforcement agencies.

“This work can provide needed updates to currently used tests, both in health care and law enforcement settings,” says Yi Xiao, associate professor of chemistry at NC State and corresponding author of two studies describing the work.

“For example, drug field testing currently used by law enforcement still relies on chemical tests developed a century ago that are poorly specific, which means they react to compounds that may not be the drug they’re looking for,” Xiao says.

“And the existing aptamer test for cocaine isn’t sensitive and specific enough to detect clinically relevant amounts of the drug in biological samples, like blood. The sensors we developed can detect cocaine in blood at nanomolar, rather than micromolar, levels, which represents a 1,000-fold improvement in sensitivity.”

Producing Hydrogen from Rocks Gains Steam as Scientists Advance New Methods

Researchers are studying chemical catalysts that can produce hydrogen gas from iron-rich rocks.
Photo Credit: Toti Larson / UT Austin.

In a project that could be a game changer for the energy transition, researchers at The University of Texas at Austin are exploring a suite of natural catalysts to help produce hydrogen gas from iron-rich rocks without emitting carbon dioxide.

If the scientists are successful, the project could jump-start a brand-new type of hydrogen industry: geologic hydrogen.

“We’re producing hydrogen from rocks,” said Toti Larson, a research associate professor at the UT Jackson School of Geosciences Bureau of Economic Geology and the lead researcher on the project. “It’s a type of non-fossil fuel production of hydrogen from iron-rich rocks that has never been attempted at an industrial scale.”

The research team recently received a $1.7 million grant from the Department of Energy and is collaborating with scientists at the University of Wyoming’s School of Energy Resources to explore the feasibility of this process on different rock types across the United States.

An evolutionary mystery 125 million years in the making

A bushel of tomatoes at the CSHL Uplands Farm.
Photo Credit: Courtesy of Cold Spring Harbor Laboratory

Plant genomics has come a long way since Cold Spring Harbor Laboratory (CSHL) helped sequence the first plant genome. But engineering the perfect crop is still, in many ways, a game of chance. Making the same DNA mutation in two different plants doesn’t always give us the crop traits we want. The question is why not? CSHL plant biologists just dug up a reason.

CSHL Professor and HHMI Investigator Zachary Lippman and his team discovered that tomato and Arabidopsis thaliana plants can use very different regulatory systems to control the same exact gene. Incredibly, they linked this behavior to extreme genetic makeovers that occurred over 125 million years of evolution.

Study of slowly evolving ‘living fossils’ reveals key genetic insights

The alligator gar, and other gar species, are “living fossils” that it shows little species diversity or physical differences from ancestors that lived tens of millions of years ago.
Photo Credit: David Solomon

In 1859, Charles Darwin coined the term “living fossils” to describe organisms that show little species diversity or physical differences from their ancestors in the fossil record. In a new study, Yale researchers provide the first evidence of a biological mechanism that explains how living fossils occur in nature.

The study, published in the journal Evolution, shows that gars — an ancient group of ray-finned fishes that fit the definition of a living fossil — have the slowest rate of molecular evolution among all jawed vertebrates, meaning their genome changes more slowly than those of other animals.

By linking this finding to the process of hybridization — when two different species produce viable offspring — of gar species in the wild that last shared common ancestry during the age of the dinosaurs, the researchers demonstrate that slow evolution rate of their genome drives their low species diversity.

“We show that gars’ slow rate of molecular evolution has stymied their rate of speciation,” said Thomas J. Near, professor of Ecology and Evolutionary Biology in Yale’s Faculty of Arts and Sciences and the paper’s senior author. “Fundamentally, this is the first instance where science is showing that a lineage, through an intrinsic aspect of its biology, fits the criteria of living fossils.” 

Study shows social factors of low U.S. Breast cancer screening

Photo Credit: Marco Jean deOliveira Teixeira

To identify major social factors hindering breast cancer screening in women aged 40 and older in the U.S., researchers focused on race/ethnicity, employment, education, food security, insurance status, housing and access to quality health care.

There is a pressing need to explore and understand which social determinants of health (SDOH) and health inequities act as significant influential factors that contribute to low breast cancer screening behaviors in the United States.

Health disparities have been consistently associated with delayed screening, which then contributes to higher mortality rates among both Hispanic and Black populations. Moreover, poverty, lack of education, neighborhood disadvantage, residential segregation, racial discrimination, lack of social support and social isolation also play a role in the breast cancer stage at diagnosis.

Researchers from Florida Atlantic University’s Schmidt College of Medicine conducted a scoping review of 72 peer-reviewed observational studies published between 2013 and 2023 to identify the major SDOH that hinder breast cancer screening in women aged 40 and older in the U.S. They focused on race/ethnicity, employment, education, food security, insurance status, housing and access to quality health care.

Low iron levels resulting from infection could be key trigger of long COVID

Photo Credit: Malachi Cowie

Problems with iron levels in the blood and the body’s ability to regulate this important nutrient as a result of SARS-CoV-2 infection could be a key trigger for long COVID, new research has discovered.

"Iron levels, and the way the body regulates iron, were disrupted early on during SARS-CoV-2 infection, and took a very long time to recover, particularly in those people who went on to report long COVID months later"

Aimee Hanson

The discovery not only points to possible ways to prevent or treat the condition, but could help explain why symptoms similar to those of long COVID are also commonly seen in a number of post-viral conditions and chronic inflammation.

Although estimates are highly variable, as many as three in 10 people infected with SARS-CoV-2 could go on to develop long COVID, with symptoms including fatigue, shortness of breath, muscle aches and problems with memory and concentration (‘brain fog’). An estimated 1.9 million people in the UK alone were experiencing self-reported long COVID as of March 2023, according to the Office of National Statistics.

Shortly after the start of the COVID-19 pandemic, researchers at the University of Cambridge began recruiting people who had tested positive for the virus for the COVID-19 cohort of the National Institute for Health and Care Research (NIHR) BioResource. These included asymptomatic healthcare staff identified via routine screening through patients admitted to Cambridge University Hospitals NHS Foundation Trust, and some to its intensive care unit.

Umbrella for Atoms: The First Protective Layer for 2d Quantum Materials

Amalgamation of experimental images. At the top, a scanning tunneling microscopy image displays the graphene’s honeycomb lattice (the protective layer). In the center, electron microscopy shows a top view of the material indenene as a triangular lattice. Below it is a side view of the silicon carbide substrate. It can be seen that both the indenene and the graphene consist of a single atomic layer.
Image Credit: © Jonas Erhardt/Christoph Maeder

As silicon-based computer chips approach their physical limitations in the quest for faster and smaller designs, the search for alternative materials that remain functional at atomic scales is one of science's biggest challenges. In a groundbreaking development, researchers at the Würzburg-Dresden Cluster of Excellence ct.qmat have engineered a protective film that shields quantum semiconductor layers just one atom thick from environmental influences without compromising their revolutionary quantum properties. This puts the application of these delicate atomic layers in ultrathin electronic components within realistic reach. The findings have just been published in Nature Communications.

2D Quantum Materials Instead of Silicon

The race to create increasingly faster and more powerful computer chips continues as transistors, their fundamental components, shrink to ever smaller and more compact sizes. In a few years, these transistors will measure just a few atoms across – by which point, the miniaturization of the silicon technology currently used will have reached its physical limits. Consequently, the quest for alternative materials with entirely new properties is crucial for future technological advancements.