Making molecules make sense: A regional explanation method reveals structure–property relationships
2025-08-01
(Press-News.org)
In cheminformatics, where machine learning is transforming our understanding of how molecular properties are predicted and explained, a critical challenge has long remained: making these powerful but often "black box" models easier to interpret. Recently, researchers at the Australian National University developed a breakthrough solution: a "regional explanation" method that helps reveal how molecular structures drive their properties. This research was published June 3 in Intelligent Computing, a Science Partner Journal, in an article titled “Regional Explanations and Diverse Molecular Representations in Cheminformatics: A Comparative Study.”
The new regional explanation method bridges the gap between local and global explanations, capturing nonlinear relationships between molecular features and properties. The authors found that different molecular representations showed consistency in their regional explanations. The new method offers fine-grained, chemically meaningful insights often missed by traditional explanation methods. It was validated on 2 datasets, demonstrating broad applicability across different chemical domains.
To develop and validate this method, the researchers chose a dataset of 2,384 graphene oxide nanoflakes, each annotated with 783 molecular features used to predict formation energy, a key indicator of thermodynamic stability. After removing duplicates, 2,116 molecules remained. The researchers tested their method on 4 different data representations of these molecules, pairing each representation with an appropriate machine learning model: a multi-layer perceptron for the tabular representation, a transformer for sequences, a convolutional neural network for images and a graph convolutional network for graph data. To ensure robust and fair comparisons, missing values were addressed, and data normalization was applied. Both local explanation methods and the regional explanation approach were used to interpret model predictions. Analysis revealed that the predictive features identified by the new approach reflected real-world knowledge about chemical properties related to formation energy. The method's generalizability was confirmed using the Quantum Machine 9 (QM9) dataset, a larger and more chemically diverse benchmark set that supports results on the real-world graphene oxide nanoflake dataset.
The researchers believe their regional explanation method could have broad application, from materials design to drug discovery, and could serve as a practical tool to understand complex structure–property relationships. Future work may focus on incorporating automated clustering to better capture property-specific molecular patterns or on adding uncertainty quantification to enhance interpretability.
END
ELSE PRESS RELEASES FROM THIS DATE:
2025-08-01
Partisan animosity is a powerful driver of protest participation -- sometimes nearly matching or even exceeding concern about the actual issues, according to a new study published in the journal Social Forces.
The research, conducted by political scientist Seth Warner at the University of Connecticut, analyzed three nationally representative surveys fielded between 2014 and 2022. Each focused on a different protest movement: Black Lives Matter, the climate movement, and the Tea Party.
Warner matched survey questions to these movements’ core grievances ...
2025-08-01
Reston, VA (August 1, 2025)—New research has been published ahead-of-print by The Journal of Nuclear Medicine (JNM). JNM is published by the Society of Nuclear Medicine and Molecular Imaging, an international scientific and medical organization dedicated to advancing nuclear medicine, molecular imaging, and theranostics—precision medicine that allows diagnosis and treatment to be tailored to individual patients in order to achieve the best possible outcomes.
Summaries of the newly published research articles are provided below.
New Imaging Tool Targets Hidden Liver Tumors
Liver cancer is hard to monitor after treatment, ...
2025-08-01
“[…] we were able to reproduce systemic rejuvenating effects of circulating blood factors on the human skin, which have been so far only demonstrated in rodent heterochronic parabiosis studies.”
BUFFALO, NY — August 1, 2025 — A new research paper featured on the cover of Volume 17, Issue 7 of Aging (Aging-US) was published on July 25, 2025, titled “Systemic factors in young human serum influence in vitro responses of human skin and bone marrow-derived blood cells in a microphysiological co-culture system.”
The study, led by first author Johanna Ritter and corresponding ...
2025-08-01
A comprehensive survey published May 23 in Intelligent Computing, a Science Partner Journal maps out the role of large language models in task planning, underscoring the growing influence of artificial intelligence in complex decision-making tasks.
Traditionally reliant on expert systems and manual configuration, task planning is essential for organizing action sequences to achieve defined goals, and is now being redefined by the advanced reasoning capabilities of large language models. The survey offers a comprehensive synthesis of how these models are reshaping ...
2025-08-01
Medicare drug plans are increasingly excluding coverage of new specialty drugs that treat complex conditions like cancers and autoimmune diseases. New research from the USC Schaeffer Center shows how these barriers may come at a cost to patients’ health.
In a large study of Medicare beneficiaries with multiple sclerosis (MS), researchers found those in plans with broader coverage of MS treatments had significantly lower risk of developing new or worsening symptoms months later. The findings, published Aug. 1 in JAMA Network Open, suggest that plans with narrower coverage of MS treatments may be linked to worse health ...
2025-08-01
A new paper published today in Cell highlights how researchers have leveraged AI-based computational protein design to create a novel synthetic ligand that activates the Notch signaling pathway, a key driver in T-cell development and function. These so-called soluble Notch agonists can be broadly applied to optimize clinical T-cell production and advance immunotherapy development.
Notch signaling is central to many cellular differentiation processes and is essential in transforming human immune cells into T-cells that target ...
2025-08-01
About The Study: This cross-sectional study found that when adolescents had their phones at school, they spent nearly an hour per school day on smartphones, with most of this time on social media. The results extend prior work indicating that smartphone use during instructional hours, especially social-specific use, is not trivial. These results also highlight developmental differences, with younger adolescents using smartphones and social media less than older adolescents.
Corresponding author: To contact the corresponding author, Scott H. Kollins, Ph.D., email scott@aura.com.
To access the embargoed study: Visit our For The Media website at this link https://media.jamanetwork.com/
(doi: ...
2025-08-01
About The Study: In this cross-sectional analysis, negative patient experiences frequently centered on quality of communication and administrative issues. Negative feedback centered on unmet expectations, whereas positive reviews emphasized supportive staff interactions.
Corresponding author: To contact the corresponding author, Neil K.R. Sehgal, M.E., email neilsehgal99@gmail.com.
To access the embargoed study: Visit our For The Media website at this link https://media.jamanetwork.com/
(doi: 10.1001/jamanetworkopen.2025.24505)
Editor’s Note: Please see the article for additional ...
2025-08-01
The earliest warning signs of multiple sclerosis (MS) may emerge more than a decade before the first classical neurological symptoms occur, according to new research from the University of British Columbia.
Published today in JAMA Network Open, the study analyzed the health records of more than 12,000 people in British Columbia and found that those with MS began using healthcare services at elevated rates 15 years before their first MS symptoms appear.
The findings challenge long-held assumptions about when the disease truly begins, offering the most comprehensive ...
2025-08-01
A new artificial intelligence (AI) tool could make it much easier—and cheaper—for doctors and researchers to train medical imaging software, even when only a small number of patient scans are available.
The AI tool improves upon a process called medical image segmentation, where every pixel in an image is labeled based on what it represents—cancerous or normal tissue, for example. This process is often performed by a highly trained expert, and deep learning has shown promise in automating this labor-intensive task.
The big challenge is that deep learning-based ...
LAST 30 PRESS RELEASES:
[Press-News.org] Making molecules make sense: A regional explanation method reveals structure–property relationships