Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Cassidy Chansirik
Coronavirus

This week’s list of data news highlights covers January 2, 2021 – January 8, 2021 and includes articles about developing new therapeutics to tackle the Ebola virus and answering questions about COVID-19 with virtual AI assistants. 

1. Repurposing Drugs with Machine Learning

Researchers at Ohio State University have used machine learning to find new uses for existing drugs. The researchers trained an algorithm to identify which drugs lower the risk of heart failure and stroke in patients who have heart disease using treatment data on 1.2 million heart disease patients, including information on their diagnostic test results and prescribed medications. The algorithm indicated that a diabetes medication called metformin and a depression and anxiety treatment called escitalopram, lowered the risk for patients with heart disease. 

2. Developing New Therapeutics for Ebola

Researchers at the University of Delaware have used supercomputer simulations to better understand how the Ebola virus shields its genetic material using a protein shell. The team sought to understand what makes the virus’s structural protein so stable by simulating the interaction of atoms in the virus. They discovered that the structural protein is helically shaped and maintains its structure by forming electrostatic interactions with the viral genetic material it protects. This knowledge is essential for developing new therapeutics against Ebola.

3. Shortening the Design Process of Materials with Machine Learning

Researchers at Sandia National Laboratories in the United States have developed a machine learning algorithm that shortens the time it takes to design materials for new technologies. Currently, the design of components grossly outpaces the design of the materials you need to build them. But with the researchers’ machine learning algorithm, scientists can complete materials science calculations 42,000 times faster than before, accelerating the design process of materials by approximately a year. 

4. Modeling the Role of Chloride in Corrosion with a Supercomputer

Researchers at Oregon State University have collaborated with the San Diego Supercomputer Center and the Texas Advanced Computing Center to model the role chloride plays in corroding materials. The team used supercomputers to model the chemical processes that lead to chloride-induced corrosion of iron to understand the structural, magnetic, and electric properties of chloride when it interacts with steel, an alloy of iron and carbon. Using this model, researchers can custom design alloys and corrosion inhibitors to increase the service life on structural metals.

5. Answering Questions about COVID-19 with AI virtual assistants

Developers at Hyro Inc., a U.S. company that develops conversational AI systems, have developed VAXA, an AI tool that diverts calls for vaccination requests and questions about COVID-19 to AI-powered virtual assistants. This lightens the load of incoming calls for call-center staff. The AI-powered virtual assistants receive 40 percent of calls and are able to converse and schedule appointments for patients. The tool also sends reminders to patients about upcoming appointments.

6. Identifying Faulty Components in Particle Accelerators with Machine Learning

Researchers at the U.S. Department of Energy have developed a machine learning tool to identify faulty components in particle accelerators, a type of machine used to explore the structure of nuclear matter. The team trained the tool to identify faulty components based on changes in radiofrequency signals that accelerated particles emit. When tested over a two-week period, the tool identified faulty components with 85 percent accuracy and the cause of the fault with 78 percent accuracy. 

7. Customizing Treatments for Patients with Oral Cancer

Researchers from Case Western University in Ohio and Vanderbilt University in Tennessee are using computer vision and machine learning to identify oral cancer, the eighth most common cancer worldwide. Researchers trained the algorithm to identify cancer and immune cells from digital images and applied computer vision to the images to recognize patterns and features among cells. With this technique, clinicians can customize treatments for patients based on how advanced their cancer is, such as conducting surgery on early-stage patients or starting chemotherapy for advanced-stage patients. 

8. Revealing the Structural Transformations of Silicon with Machine Learning

Researchers at the University of Oxford have used machine learning to model how the behavior of silicon atoms change as they move from liquid states, where they are electrical conductors, to solid states, where they are semiconducting only in ambient conditions—a fact that underpins silicon use in technologies ranging from computer chips to solar panels. Simulations revealed that structural transformations of silicon atoms do not occur simultaneously under pressure, but rather evolve gradually, a finding that developers can use to inform how silicon is manipulated in electronic communication technologies and energy harvesting. 

9. Finding Potential COVID-19 Treatments with a Supercomputer

Researchers at the San Diego Supercomputer Center have used a supercomputer to identify drugs approved for other diseases that might work as treatments for COVID-19. Using molecular simulations, researchers found 147 compounds, including Vitamin D3 and calcium glubionate, a mineral supplement found in citrus fruits, that offer promising potential treatments. Currently, a team of researchers at Harvard University are conducting a nationwide study on Vitamin D3 as a treatment for COVID-19. 

10. Using AI to Better Produce Images from Text

Scientists at OpenAI, a U.S. company that conducts AI research, have developed DALL-E, a neural network that can produce original images based on short written descriptions. Unlike previous text-to-image programs, DALL-E can infer details not explicitly mentioned in the description but would be required for a realistic image. For example, given the description “a painting of a fox sitting in a field during winter,” DALL-E deduced the need to include a shadow in the picture. DALL-E can also combine multiple objects, provide different points of views, and generate images from novel and unrelated concepts. 

Image credit: U.S. Centers for Disease Control and Prevention 

You may also like

Show Buttons
Hide Buttons