Cambridge develops IcePic algorithm that outperforms humans in predicting ice crystal formation

Cambridge scientists have developed an artificially intelligent algorithm capable of beating scientists at predicting how and when different materials form ice crystals.

The program – IcePic – could help atmospheric scientists improve climate change models in the future. 

Water has some unusual properties, such as expanding when it turns into ice. Understanding water and how it freezes around different molecules have wide-reaching implications in a broad range of areas, from weather systems that can affect whole continents to storing biological tissue samples in a hospital.

The Celsius temperature scale was designed based on the premise that it is the transition temperature between water and ice; however, whilst ice always melts at 0°C, water doesn’t necessarily freeze at 0°C. Water can still be in liquid form at -40°C, and it is impurities in water that enable the ice to freeze at higher temperatures. One of the biggest aims of the field has been to predict the ability of different materials to promote the formation of ice – known as a material’s “ice nucleation ability”.

Researchers at the University of Cambridge, have developed a ‘deep learning’ tool able to predict the ice nucleation ability of different materials – and which was able to beat scientists in an online ‘quiz’ in which they were asked to predict when ice crystals would form.

Deep learning is how artificial intelligence (AI) learns to draw insights from raw data. It finds its patterns in the data, freeing it of the need for human input so that it can process results faster and more precisely. In the case of IcePic, it can infer different ice crystal formation properties around different materials. IcePic has been trained on thousands of images so that it can look at completely new systems and infer accurate predictions from them.

The team set up a quiz in which scientists were asked to predict when ice crystals would form in different conditions shown by 15 different images. These results were then measured against IcePic’s performance. When put to the test, IcePic was far more accurate in determining a material’s ice nucleation ability than over 50 researchers from across the globe. Moreover, it helped identify where humans were going wrong.

Michael Davies, a Ph.D. student in the ICE lab at the Yusuf Hamied Department of Chemistry, Cambridge, and University College London, London, the first author of the study, said: “It was fascinating to learn that the images of water we showed IcePic contain enough information to predict ice nucleation.

“Despite us – that is, human scientists – having a 75-year head start in terms of the science, IcePic was still able to do something we couldn’t.”

Determining the formation of ice has become especially relevant in climate change research.

Water continuously moves within the Earth and its atmosphere, condensing to form clouds and precipitating in the form of rain and snow. Different foreign particles affect how ice forms in these clouds, for example, smoke particles from pollution compared to smoke particles from a volcano. Understanding how different conditions affect our cloud systems is essential for more accurate weather predictions.

“The nucleation of ice is really important for the atmospheric science community and climate modeling,” said Davies. “At the moment there is no viable way to predict ice nucleation other than direct experiments or expensive simulations. IcePic should open up a lot more applications for discovery.” 

UCR astrophysicist shows why Jupiter doesn’t have rings like Saturn

Gigantic moons prevent giant icy ring formations

Because it’s bigger, Jupiter ought to have larger, more spectacular rings than Saturn has. But new UC Riverside research shows Jupiter’s massive moons prevent that vision from lighting up the night sky.

“It’s long bothered me why Jupiter doesn’t have even more amazing rings that would put Saturn’s to shame,” said UCR astrophysicist Stephen Kane, who led the research.  Artist rendering of Jupiter with rings that rival Saturn's. (Stephen Kane/UCR)

“If Jupiter did have them, they’d appear even brighter to us, because the planet is so much closer than Saturn.” Kane also had questions about whether Jupiter once had fantastic rings and lost them. Ring structures can be temporary.

To understand the reason Jupiter currently looks the way it does, Kane and his graduate student Zhexing Li ran a dynamic supercomputer simulation accounting for the orbits of Jupiter’s four main moons, as well as the orbit of the planet itself, and information about the time it takes for rings to form. Their results are now online, soon to be published in the Planetary Science journal.

Saturn’s rings are largely made of ice, some of which may have come from comets, which are also largely made of ice. If moons are massive enough, their gravity can toss the ice out of a planet’s orbit, or change the orbit of the ice enough so that it collides with the moons.

“We found that the Galilean moons of Jupiter, one of which is the largest moon in our solar system, would very quickly destroy any large rings that might form,” Kane said. As a result, it is unlikely that Jupiter had large rings at any point in its past.

“Massive planets form massive moons, which prevents them from having substantial rings,” Kane said. 

All four giant planets in our solar system — Saturn, Neptune, Uranus, and also Jupiter — do have rings. However, both Neptune and Jupiter’s rings are so flimsy they’re difficult to view with traditional stargazing instruments. 

Coincidentally, some of the recent images from the newly commissioned James Webb Space Telescope included pictures of Jupiter, in which the faint rings are visible.

“We didn’t know these ephemeral rings existed until the Voyager spacecraft went past because we couldn’t see them,” Kane said. 

Uranus has rings that aren’t as large but are more substantial than Saturn’s. Going forward, Kane intends to run simulations of the conditions on Uranus to see what the life of that planet’s rings might be. 

Some astronomers believe Uranus is tipped over on its side as the result of a collision the planet had with another celestial body. Its rings could be the remains of that impact. 

Beyond their beauty, rings help astronomers understand the history of a planet because they offer evidence of collisions with moons or comets that may have happened in the past. The shape and size of the rings, as well as the composition of the material, offer an indication of the type of event that formed them.

“For us astronomers, they are the blood spatter on the walls of a crime scene. When we look at the rings of giant planets, it’s evidence something catastrophic happened to put that material there,” Kane said. 

UNC reveals genetics underlying alcohol, cigarette abuse

UNC School of Medicine researchers led by Hyejung Won, Ph.D., used a new kind of computational tool to parse the complicated genetics that put some people at higher risk of becoming addicted to alcohol, cigarettes, and likely other substances of abuse. Hyejung Won, PhD

Have you ever wondered why one person can smoke cigarettes for a year and easily quit, while another person will become addicted for life? Why can’t some people help themselves from abusing alcohol and others can take it or leave it? One reason is a person’s genetic proclivity to abuse substances. UNC School of Medicine researchers led by Hyejung Won, Ph.D., are beginning to understand these underlying genetic differences. The more they learn, the better chance they will be able to create therapies to help the millions of people who struggle with addiction.

Won, assistant professor of genetics and member of the UNC Neuroscience Center, and colleagues identified genes linked to cigarette smoking and drinking. The researchers found that these genes are over-represented in certain kinds of neurons – brain cells that trigger other cells to send chemical signals throughout the brain.

The researchers, who published their work in the journal Molecular Psychiatry, also found that the genes underlying cigarette smoking were linked to the perception of pain and response to food, as well as the abuse of other drugs, such as cocaine. Other genes associated with alcohol use were linked to stress and learning, as well as abuse of other drugs, such as morphine.

Nancy Sey

Given the lack of current treatment options for substance use disorder, the researchers also conducted analyses of a publicly available drug database to identify potential new treatments for substance abuse.

“We found that antipsychotics and other mood stabilizers could potentially provide therapeutic relief for individuals struggling with substance abuse,” said first author Nancy Sey, a graduate student in the Won lab. “And we’re confident our research provides a good foundation for research focused on creating better treatments to address drug dependency.”

Parsing the Genome

Long-term substance use and substance use disorders have been linked to many common diseases and conditions, such as lung cancer, liver disease, and mental illnesses. Yet, few treatment options are available, largely due to gaps in our understanding of the biological processes involved.

“We know from twin studies that genetics may account for why some people use and abuse substances, aside from environmental factors, such as family issues or personal trauma,” Won said. “Genetic studies such as genome-wide association studies (GWAS) provide a way to identify genes associated with complex human traits, such as nicotine addiction or drinking heavily.”

Through GWAS, Won added, researchers can identify regions in the genome that play roles in particular traits, compared to individuals who do not exhibit the trait. Yet, genome-wide studies cannot tell us much about how genes in those regions affect a trait. That’s because these regions are often in “non-coding” regions of the genome.

“Non-coding” refers to the fact that the genes in these regions do not translate – or code – their genetic information directly into the creation of proteins, which then perform a known biological function. Therefore, what actually happens biologically in these “non-coding” regions remains mostly unknown.

“We wanted to learn what’s happening in these regions,” Won said. “So we developed Hi-C coupled MAGMA (H-MAGMA), a computational tool to help us make more sense of what we’re seeing in genome-wide studies.”

In a previous publication, Won’s lab showed how applying H-MAGMA to brain disorders identifies their associated genes and described their underlying biology. And for this current paper, her lab expanded the tool to cigarette smoking and drinking.

They developed H-MAGMA frameworks from dopaminergic neurons and cortical neurons – brain cell types that researchers have long implicated in substance use. Focusing on those two cell types, Won’s team – led by Sey, an HHMI Gilliam Fellow – applied H-MAGMA to GWAS findings related to heaviness of smoking, nicotine dependence, problematic alcohol use, and heaviness of drinking to identify genes associated with each trait.

Genes associated with alcohol use and cigarette smoking were also associated with other types of substances, such as morphine and cocaine. While the opioid crisis has caused a detrimental social burden, well-powered GWAS on cocaine and opioid use is not currently available. Won’s team, therefore, sought to determine whether the genes associated with alcohol use and cigarette smoking can reveal genetics underlying general addiction behavior, genetic findings that could be extended to other substances of abuse.

“Our analyses showed that expression of genes shared between cigarette smoking and alcohol use traits can be altered by other types of substances such as cocaine,” Won said. “By characterizing the biological function of these genes, we will be able to identify the biological mechanisms underlying addiction, which could be generalized to various forms of substance use disorder.”

In addition to the various types of excitatory neurons, Won’s team also identified additional cell types, including cortical glutamatergic, midbrain dopaminergic, GABAergic, and serotonergic neurons that are associated with the risk genes.

With these findings in hand, it is now possible for UNC researchers and others to investigate molecules that make addiction much less likely.