Kazan University chemists teach neural networks to predict properties of compounds

A new joint Russian-French-Japanese paper appeared in the Journal of Chemical Information and Modeling

The international team works on a computational model able to predict the properties of new molecules based on the analysis of fundamental chemical laws. The project was supported by the Russian Science Foundation (title "Using AI methods for the planning of chemical synthesis").

Co-author, Associate Professor Timur Madzhidov, explains, "We offered a way to insert the preexisting chemical equations into some frameworks of machine learning. It was tested on the predictions of tautomeric constants and acidity, which are linked by the Kabachnik equation. Using the functional interdependency between them, the neural network learns how to predict both these properties." {module INSIDE STORY}

Prototropic tautomerism is the phenomenon of reversible isomerism, in which isomers (substances having the same qualitative and quantitative composition, but differing in structure and properties) easily transition into each other due to the transfer of a hydrogen atom.

"Tautomeric transformations are very common for organic compounds, are known for about half of all discovered compounds. For example, one of the mechanisms of spontaneous mutations is tied to the tautomeric transformations of DNA nucleic base. That why tautomerism must be taken into account when registering new compounds, during the computer design of new medications and the search for molecules with preconditioned properties," adds Madzhidov.

The results of this research can help increase the precision of the prediction of physicochemical properties of designed medication and materials, as well as correctly forecast the parameters of chemical reactions.

Kazan Federal University, Lomonosov Moscow State University, the University of Hokkaido, and the University of Strasbourg contributed to the publication.

University of Houston researchers solve a scientific mystery about evaporation implications for power generation, desalination, electronics

Evaporation can explain why water levels drop in a full swimming pool, but it also plays an important role in industrial processes ranging from cooling electronics to power generation. Much of the global electricity supply is generated by steam plants, which are driven by evaporation.

But determining when and how quickly a liquid will convert to vapor has been stymied by questions about how - and how much - the temperature changes at the point where the liquid meets the vapor, a concept known as temperature discontinuity. Those questions have made it more difficult to create more efficient processes using evaporation, but now researchers from the University of Houston have reported answers to what happens at that interface, addressing 20 years of conflicting findings. The work was reported in the Journal of Physical ChemistryHadi Ghasemi, Cullen Associate Professor of Mechanical Engineering at the University of Houston, led research that eliminates the {module INSIDE STORY}

The temperature discontinuity was first reported in 1999 by Canadian researchers G. Fang and C.A. Ward, who noted that they were unable to explain the phenomenon through classical mechanics. The new work solves that mystery.

Hadi Ghasemi, Cullen Associate Professor of Mechanical Engineering at UH, said the new understanding eliminates the "bottleneck" that has complicated predictions and simulations of processes involving evaporation.

"We demonstrated the physics of what happens within the space of a few molecules at the interface and accurately developed a theory on the evaporation rate," Ghasemi said. "That allowed us to explain all of the conflicting findings that have been reported in the last 20 years and solve this mystery."

In addition to Ghasemi, co-authors for the paper included first author Parham Jafari, a Ph.D. student at UH, and Amit Amritkar, a research assistant professor at UH.

The researchers first approached the question in the lab, but Ghasemi said they were unable to get the needed spatial resolution for a definitive answer. They used a computational approach in order to find the properties of liquid and vapor within the length of a few molecules.

The explanation - developed using the Direct Simulation Monte Carlo method - will allow scientists to more accurately simulate the performance of all systems based on the theory of evaporation.

"With this understanding, we can more accurately develop simulations of performance and efficiency, as well as design and predict the behavior of advanced systems," Ghasemi said.

That would have applications for energy, electronics, photonics, and other fields.

As just one example of the importance of evaporation, Ghasemi noted that 80% of electric power globally is generated through steam plants, which work based on evaporation phenomena.

University of Cincinnati geographers find tipping point in deforestation

Global satellite imagery shows transitions occur quickly after blocks of forest are cut in half

University of Cincinnati geography researchers have identified a tipping point for deforestation that leads to rapid forest loss.

Geography professor Tomasz Stepinski used high-resolution satellite images from the European Space Agency to study landscapes in 9-kilometer-wide blocks across every inch of the planet between 1992 and 2015. He found that deforestation occurs comparatively slowly in these blocks until about half of the forest is gone. Then the remaining forest disappears very quickly. A University of Cincinnati land-use map shows changing landscapes in North and South America between 1992 and 2015. White indicates little or no change. Darker shades indicate the highest rate of change in each category. Forest loss was the most noticeable category in Central and South America.{module INSIDE STORY}

The study was published in the journal Geophysical Research Letters.

Stepinski and former UC postdoctoral researcher Jakub Nowosad, the lead author, discovered something surprising and fundamental: nature abhors mixed landscapes, at least on a scale of 81 square kilometers. The study showed that mixed landscapes (like agriculture and forest) are comparatively few and, more surprisingly, do not stay mixed for long. These mixed blocks tend to become homogeneous over time, regardless of the landscape type.

"I think it's very intuitive. It corresponds to the different climatic zones. The Earth before people was certainly like that. You had forests and mountains and wetlands and deserts," Stepinski said. "You would expect people would create more fragmentation, but as it turns out, people never stop. They convert the entire block on a large scale."

Stepinski said landscapes are always changing through natural or anthropological causes. Human causes are both direct, like clear-cutting, or indirect like climate change.

Last year, Stepinski used the same data to demonstrate that 22% of the Earth's habitable surface was altered in measurable ways between 1992 and 2015. The biggest change: forest to agriculture.

For the new study, Stepinski examined nearly 1.8 million blocks covering Earth's seven continents. Blocks were categorized by 64 landscape combinations. Researchers observed transitions in these blocks from predominantly one type to predominantly another in nearly 15% of the blocks between 1992 and 2015.

"The data we have covers 23 years. That's a relatively short period of time. But from that we can calculate change in the future," Stepinski said.

Deforestation was the most pronounced example of human-caused landscape change, researchers found. They used probability modeling known as Monte Carlo methods to determine the likelihood of different types of landscape change over time (in this case hundreds of years).

The result? Researchers found that the most likely trajectory of change was from one homogeneous type to another.

"Planet Earth wants to be homogeneous. The land wants to be the same in all these patches. And when they start to change, they don't stop until they convert everything into another homogeneous block," he said.

The authors did not examine why blocks change so quickly once a transition begins. But Stepinski said it's possible that development such as logging roads or drainage required to clear forest makes continued change that much easier.

"I can only speculate because that was not part of the study, but I would imagine two things are happening," he said. "If you are cutting forest, you have the infrastructure to finish it. It's so much easier to cut the rest. Second, the forest is more vulnerable to change when there has been a disturbance." 

Wildlife managers often try to preserve larger intact blocks to prevent fragmentation, said Martin McCallister, the Appalachian Forest project manager for the Edge of Appalachia Nature Preserve in southern Ohio. The preserve is managed by the Nature Conservancy, one of the world's largest wildlife conservation organizations.

"You'd be hard pressed to find land managers who wouldn't be strongly in favor of protecting larger tracts because they're more resilient to a variety of challenges, including invasive species and climate change," McCallister said. "Once a property gets fragmented by roads, it's easier to extract resources. It's also easier for invasive species and pests to get a foothold."

McCallister said woodlands can be fragmented on paper, too.

"In Ohio, 96 percent of our woodland owners have less than 50 acres. They represent a lot of small parcels," he said.

The UC study found that mixed land types don't stay mixed for long. 

"I think it is interesting that this property applies both to natural and human landscapes," said co-author Nowosad, a former UC postdoctoral researcher who now works as an assistant professor at the Adam Mickiewicz University in Poland.

Nowosad said the study provides a data-driven model of long-term landscape change. While researchers only looked at changes between forest and agriculture, Nowosad said it would be worthwhile to examine whether tipping points exist for other landscape transitions.

"This model can be used to help understand how landscapes evolved and are going to evolve in the future," Nowosad said.

Stepinski, a physicist who worked for NASA before coming to UC, said the principle borrows from other disciplines, particularly astrophysics.

"If you look at the evolution of stars, the principle is you predict a long-term path statistically from short-term knowledge," Stepinski said. "It's an idea that has been used elsewhere but never for environmental study."

While it's only a theory, it's one that could be borne out by time, he said.

"It's thought-provoking. My hope is that people will criticize it and come up with different ideas," Stepinski said.