Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
Researchers found that many changes to human DNA had opposing effects, with some variants making enhancers stronger while others made them weaker. This discovery has implications for understanding human evolution and the potential link between human DNA variations and psychiatric diseases.
A research team at Carnegie Mellon University has developed a machine learning method called SPICEMIX to analyze spatial transcriptomics data. The tool helps identify and understand gene expression patterns in cells, revealing new insights into brain cell types.
Researchers at Brookhaven National Laboratory have successfully discovered new materials using artificial intelligence and self-assembly. The AI-driven technique led to the discovery of three new nanostructures, expanding the scope of self-assembly's applications in microelectronics and catalysis.
Researchers developed machine learning models to accurately calculate fine particulate matter in urban air pollution using AI and traffic data. The models provide a high-resolution estimation of city street pollution surface, enabling transportation and epidemiology studies to assess health impacts.
A large randomized study found that machine learning-triggered reminders significantly increased rates of advanced care planning conversations, reducing potentially harmful therapies at end of life. The intervention also improved patient education and early palliative care referrals.
Researchers designed a new long-acting injectable drug formulation using machine learning algorithms, achieving a slow-release rate in just one iteration. The study demonstrates the potential for machine learning to accelerate the development of innovative drug delivery technologies.
A new study reveals that high-quality coral reefs in Hawaii are popular tourist sites, but also at risk from tourism-related development and pollution. The research used social media and aerial mapping to analyze the impact of tourist visitation on live coral cover across hundreds of coastal sites.
A new study uses machine learning to predict poor glycemic control in patients with type 2 diabetes, identifying key factors such as prior glucose levels and anti-diabetic medicines. The findings suggest that data routinely collected for diabetes monitoring can reliably identify patients at risk of hyperglycemia.
Researchers have developed a diffractive optical processor that can compute hundreds of transformations in parallel using wavelength multiplexing. The processor, which is powered by light instead of electricity, can execute multiple complex functions simultaneously at the speed of light.
Researchers used machine learning to create molecule chains that display designated colors in response to different stimuli, such as light, chemicals, and energy. This breakthrough enables faster and more efficient data storage and security applications.
A UVA research team developed a real-time detection method for keyhole pore generation in laser powder bed fusion, achieving a 100% prediction rate. This approach expands additive manufacturing capabilities for aerospace and other industries relying on strong metal parts.
A group of scientists developed a machine learning approach to predict amine emissions from a carbon capture plant. They analyzed data from a stress test at a German power plant and found that two amines respond in opposite ways, increasing or decreasing emissions. This new method has the potential to change the way chemical plants ope...
Researchers develop AI model to predict exhaust gas emissions from ships under different air-to-fuel ratios. The ensemble dataset and double ensemble models produce the most accurate emission predictions for CO2, NOx, and SO2 gases.
Scientists have developed a new method to enhance electron-photon coupling, resulting in a hundredfold increase in light emissions. The approach uses a specially designed photonic crystal to produce stronger interactions between photons and electrons.
ETH Zurich researchers have created a range of affordable fluorescent inks with machine learning algorithms to determine the right molecular subunits. The new dyes can be used for security features and applications like solar power plants and organic light-emitting diodes.
A new computer program, DeepMosaic, uses artificial intelligence to detect mosaic mutations in genetic sequences. This method enables accurate detection of mosaic mutations, which cause hundreds of unsolved and untreatable disorders, including epilepsy.
Researchers at Drexel University used GPT-3 to spot early signs of Alzheimer's in spontaneous speech, achieving 80% accuracy. The program analyzed word-use, sentence structure and meaning from transcripts to identify characteristic profiles of Alzheimer's speech.
Researchers at University of Copenhagen developed a method to map individual trees' carbon content using aerial images, improving accuracy and enabling better comparisons between countries. The method supports Rwanda in verifying commitments under schemes like REDD+ and AFR 100.
Scientists at Caltech used machine-learning algorithm to chart sills, mapping them with precision and linking them to active volcanoes Mauna Loa and Kīlauea. The study provides new insights into magma storage and transport deep beneath Hawai‘i.
Scientists from CHOP and NJIT created a software tool to analyze information from a single cell, revealing relationships between different cellular characteristics. The 'single-cell multimodal deep clustering' method can help identify the causes of genetic-based diseases by integrating data on gene expression, mRNA, proteins, and organ...
Researchers found that providing language descriptions of tools can accelerate a simulated robotic arm's learning of tool manipulation. The team used GPT-3 to obtain tool descriptions and showed improved performance in tasks such as pushing, lifting, sweeping, and hammering with new tools.
A digital detection of dementia study aims to evaluate the practical use of an AI tool for early identification of Alzheimer's disease and related dementias. The study uses a passive digital marker and patient-reported outcomes survey, which have shown potential in detecting mild cognitive impairment and reducing healthcare costs.
A recent study has identified common and unique cellular processes in six neurodegenerative diseases, providing new insights into the underlying causes of these conditions. The research used machine learning analysis to compare RNA markers in whole blood samples from patients with distinct diseases, revealing eight shared themes across...
A new machine learning model developed by Aalto University researchers can identify small molecules with unprecedented accuracy, distinguishing between mirror image molecules. This breakthrough has significant implications for understanding metabolic disorders, such as diabetes, and identifying micropollutants in the environment.
Researchers from Brown and MIT developed a new framework that uses machine learning and sequential sampling to predict rare disasters like earthquakes and pandemics with less data. The framework, called DeepOnet, has been shown to outperform traditional modeling efforts in predicting scenarios, probabilities and timelines of rare events.
Researchers are conducting on-site surveys and generating high-resolution damage maps for 20-square-mile region affected by the Category 4 storm. The goal is to inform protection efforts and help communities recover from the disaster.
Using machine learning to study water's phase changes, researchers found strong computational evidence in support of liquid-liquid transition. This technique can be applied to real-world systems that use water, informing water's use in industrial processes and climate models.
The UTEP-led Computing Alliance will receive $4.8M from Google to improve diversity in computer science fields. The project aims to attract, prepare and support Hispanic students for graduate degrees, with initiatives including lab design, financial support and research collaborations.
Researchers developed AI-powered software to measure and classify pacu fish, enabling breeders to select animals with higher fillet yield and faster weight gain. The system uses deep learning and can recognize different parts of the fish even in challenging environments.
Veterinary experts warn that AI algorithms used in radiology and imaging can provide faulty or incomplete diagnoses, posing risks to patient care. The absence of regulatory oversight for veterinary AI products increases concerns over accuracy, transparency, and potential harm.
Researchers developed a machine learning model that uses microbiome data from wastewater to estimate the number of individuals represented in a sample. The method was trained on over 1,100 people's samples and can be used to link wastewater properties to individual-level data.
Researchers developed a machine learning-based strategy for direct classification of acute aquatic toxicity, explaining 90% of training set variance and 80% of test set variance. The approach resulted in a fivefold decrease in incorrect categorization compared to QSAR regression models.
A new study found that a machine learning-based blood test can accurately predict an individual's entire diet over 19 food groups, outperforming traditional methods. The test, which uses molecular profiling, also identifies who is more likely to develop diabetes and cardiovascular disease based on each food group.
Using supercomputers and machine learning, researchers created simulations of millions of computer-generated universes to test astrophysical predictions. The study found that supermassive black holes grow in the same way as their host galaxies, revealing a long-elusive relationship.
Researchers developed an accurate predictive model to distinguish between COVID-19 positive and negative test outcomes. The study found that symptom features, such as fever and difficulty breathing, play a significant role in predicting test results, with molecular tests yielding lower positive rates due to their dependence on viral load.
Researchers have identified a method that can automatically detect doxing on Twitter with high accuracy, which could help protect users from cyberbullying. The approach uses machine learning to differentiate between self-disclosures and malicious disclosures of sensitive personal information.
A team of researchers from Tokyo University of Science developed a super-hierarchical and explanatory analysis method for magnetic reversal processes, enabling the detection of subtle microscopic changes. The new algorithm can predict stable/metastable states in advance and improve the reliability of spintronics devices.
A Penn State-led research team found that TikTok's unique interface and algorithm make content creation and virality particularly easy yet lead to high rates of creator burnout. The platform's central role of the algorithm in determining viewership makes creators produce new content continuously, leading to burnout.
A University of Houston researcher has developed a method to describe complex systems using the least number of variables possible, reducing complexity from millions to just one. This advancement speeds up science with efficiency and ability to understand and predict natural system behavior.
Researchers at UCSF and IBM Research create a predictive model that encodes commands for cells to kill cancer cells. By combining words that guide engineered immune cells, they can predict which elements should be included in a cell to carry out precise behaviors. This advance allows scientists to rapidly design new cellular therapies.
Researchers aim to provide the highest quality of care possible by understanding the sound environment of child care centers. A 48-hour monitoring period and staff evaluations helped identify factors influencing child and provider experiences.
Researchers have developed a scaled-up version of a probabilistic computer using stochastic spintronic devices, suitable for combinatorial optimization and machine learning. The new design combines conventional semiconductor chips with modified spintronic devices, achieving massive improvements in throughput and power consumption.
The National Cancer Institute awards $10.5 million to USC's Division of Biostatistics to develop statistical methods for uncovering new risk factors associated with cancer by integrating large volumes of health, genomic, and exposure data. The project aims to provide new insights into complex biological processes and discoveries of nov...
Researchers developed a noninvasive microphone sensor that uses machine learning to detect bowel diseases like cholera. The algorithm analyzes audio data from toilet sounds, identifying consistent tones for urination and singular tones for defecation.
Machine learning is being explored as a tool to speed up the identification of biomaterials. Researchers have also developed a guide on how to incorporate ML into research programs. Additionally, studies have investigated ways to model polymers at multiple scales and created a self-healing hydrogel for sustained release of medications.
A new AI evaluation framework, GOPHER, has been developed to assess the efficiency of genome analysis algorithms. The tool judges programs on their ability to learn genomic biology, predict patterns, handle noise, and provide interpretable decisions.
Researchers developed a machine learning algorithm to identify cough sounds and determine if someone has pneumonia, aided by room impulse responses. The algorithm can work in any environment, facilitating non-face-to-face treatment and reducing medical costs.
A team of researchers from the University of Pennsylvania has developed a new algorithm, metadynamics, that can navigate high-dimensional energy landscapes to find low-energy configurations. This breakthrough has the potential to revolutionize fields such as protein folding and machine learning.
A new paper by researchers at the University of Birmingham argues that a 'one size fits all' approach to treating early psychosis may not be effective. Instead, they propose using machine learning techniques to deliver tailored treatment plans that address individual needs and improve outcomes.
Researchers at Klick Applied Sciences have created a machine learning model to predict diabetes onset in patients using just 12 hours of data from continuous glucose monitors. The study showed high accuracy in identifying prediabetes, healthy patients, and those with Type 2 diabetes, offering a potential tool for early disease prevention.
Researchers developed a new approach to analyze coercivity in soft magnetic materials using machine learning and data science. The method condenses relevant information from microscopic images into a two-dimensional feature space, visualizing the energy landscape of magnetization reversal. This study showcases how materials informatics...
A study led by Kyoto University researchers found that AI-generated haiku poems, created without human intervention, were often indistinguishable from those penned by humans. In contrast, human-AI collaboration produced more creative works.
Researchers developed a method to measure overall fitness using wearable devices, outperforming current consumer smartwatches and fitness monitors. The model uses machine learning to predict VO2max during everyday activity, providing accurate predictions based on heart rate and accelerometer data.
Scientists used AI-driven PandaOmics platform to analyze gene expression datasets from DNA repair diseases, identifying biomarkers associated with treatment response. The study focused on genes that stratify cancer patients by survival outcomes, providing potential targets for personalized therapies.
Researchers developed a new method to detect lung cancer using machine learning and statistical techniques, identifying 7 specific volatile organic compounds in breath samples. This breakthrough could lead to earlier diagnosis and improved treatment outcomes for patients.
A new AI method has analyzed substance use trends among Canadian high schoolers, identifying factors such as large weekly allowances and low physical activity that increase the risk of transitioning to multiple substance use. The study found that once students start using substances, it is rare for them to stop, highlighting the need f...
A new machine learning fusion model has been developed to diagnose ovarian cancer more accurately by combining ultrasound and photoacoustic tomography imaging. The model achieved an accuracy of 90% in detecting ovarian lesions, outperforming previous methods.
A recent study published in Nature Computational Science reveals that specific regions of the brain process both individual and combined words, while others focus solely on individual words. The research could contribute to the development of wearable neurotechnology devices that can decode language directly from brain activity.
A neural network trained using a diverse dataset outperforms conventionally trained algorithms by reducing bias in artificial intelligence. The use of images from low-resource populations boosts the object recognition performance of machine learning systems.
The researchers have developed an AI algorithm called M3GNet that can predict the structure and dynamic properties of any material. The algorithm was used to create a database of over 31 million yet-to-be-synthesized materials with predicted properties, facilitating the discovery of new technological materials.