Lifeboat Foundation

The Human Cell Atlas is the world’s largest, growing single-cell reference atlas. It contains references of millions of cells across tissues, organs and developmental stages. These references help physicians to understand the influences of aging, environment and disease on a cell—and ultimately diagnose and treat patients better. Yet, reference atlases do not come without challenges. Single-cell datasets may contain measurement errors (batch effect), the global availability of computational resources is limited and the sharing of raw data is often legally restricted.

Researchers from Helmholtz Zentrum München and the Technical University of Munich (TUM) developed a novel algorithm called “scArches,” short for single-cell architecture surgery. The biggest advantage: “Instead of sharing raw data between clinics or research centers, the algorithm uses transfer learning to compare new datasets from single-cell genomics with existing references and thus preserves privacy and anonymity. This also makes annotating and interpreting of new data sets very easy and democratizes the usage of single-cell reference atlases dramatically,” says Mohammad Lotfollahi, the leading scientist of the algorithm.

Russian start-up NTechLab has released FindFace Multi, a detection technology that uses an advanced algorithm to recognize not only faces, but also bodies of people and cars. This is an update to the company’s flagship product and is able to support numerous video streams and facial database entries.

Body recognition allows FindFace Multi users to count and search people moving through an environment as well as identifying individuals and tracking movements. The algorithm also takes into account markers such as height, color of clothes and accessories.

The vehicle recognition function determines the body type, color, manufacturer, and model of a car, as well as searching by license plate. Even if license plates, or parts of the vehicle are not visible or obscured, the system can still identify a car.

Wide Area Networks (WANs), the global backbones and workhorses of today’s internet that connect billions of computers over continents and oceans, are the foundation of modern online services. As COVID-19 has placed a vital reliance on online services, today’s networks are struggling to deliver high bandwidth and availability imposed by emerging workloads related to machine learning, video calls, and health care.

To connect WANs over hundreds of miles, fiber optic cables that transmit data using light are threaded throughout our neighborhoods, made of incredibly thin strands of glass or plastic known as optical fibers. While they’re extremely fast, they’re not always reliable: They can easily break from weather, thunderstorms, accidents, and even animals. These tears can cause severe and expensive damage, resulting in 911 service outages, lost connectivity to the internet, and inability to use smartphone apps.

Continue reading “ARROW, a reconfigurable fiber optics network, aims to take on the end of Moore’s law” »

“The network learned to find fundamental concepts that are key to molecular structure formation, but without explicitly being told to,” Townshend added. “The exciting aspect is that the algorithm has clearly recovered things that we knew were important, but it has also recovered characteristics that we didn’t know about before.”

Having shown success with proteins, the researchers turned their attention to RNA molecules. The researchers tested their algorithm in a series of “RNA Puzzles” from a longstanding competition in their field, and in every case, the tool outperformed all the other puzzle participants and did so without being designed specifically for RNA structures.

“We introduce a machine learning approach that enables identification of accurate structural models without assumptions about their defining characteristics, despite being trained with only 18 known RNA structures,” the authors of the Science article wrote. “The resulting scoring function, the Atomic Rotationally Equivariant Scorer (ARES), substantially outperforms previous methods and consistently produces the best results in community-wide blind RNA structure prediction challenges.”

Analog photonic solutions offer unique opportunities to address complex computational tasks with unprecedented performance in terms of energy dissipation and speeds, overcoming current limitations of modern computing architectures based on electron flows and digital approaches.

In a new study published on August 26 2021, in the journal Nature Communications Physics, researchers led by Volker Sorger, an associate professor of electrical and computer engineering at the George Washington University, reveal a new nanophotonic analog processor capable of solving partial differential equations. This nanophotonic processor can be integrated at chip-scale, processing arbitrary inputs at the speed of light.

The research team also included researchers at the University of California, Los Angeles, and City College of New York.

An international research team with participants from several universities including the FAU has proposed a standardized registry for artificial intelligence (AI) work in biomedicine to improve the reproducibility of results and create trust in the use of AI algorithms in biomedical research and, in the future, in everyday clinical practice. The scientists presented their proposal in the journal Nature Methods.

In the last decades, new technologies have made it possible to develop a wide variety of systems that can generate huge amounts of biomedical data, for example in cancer research. At the same time, completely new possibilities have developed for examining and evaluating this data using artificial intelligence methods. AI algorithms in intensive care units, e.g., can predict circulatory failure at an early stage based on large amounts of data from several monitoring systems by processing a lot of complex information from different sources at the same time, which is far beyond human capabilities.

This great potential of AI systems leads to an unmanageable number of biomedical AI applications. Unfortunately, the corresponding reports and publications do not always adhere to best practices or provide only incomplete information about the algorithms used or the origin of the data. This makes assessment and comprehensive comparisons of AI models difficult. The decisions of AIs are not always comprehensible to humans and results are seldomly fully reproducible. This situation is untenable, especially in clinical research, where trust in AI models and transparent research reports are crucial to increase the acceptance of AI algorithms and to develop improved AI methods for basic biomedical research.

Determining the 3D shapes of biological molecules is one of the hardest problems in modern biology and medical discovery. Companies and research institutions often spend millions of dollars to determine a molecular structure—and even such massive efforts are frequently unsuccessful.

Using clever, new machine learning techniques, Stanford University Ph.D. students Stephan Eismann and Raphael Townshend, under the guidance of Ron Dror, associate professor of computer science, have developed an approach that overcomes this problem by predicting accurate structures computationally.

Most notably, their approach succeeds even when learning from only a few known structures, making it applicable to the types of molecules whose structures are most difficult to determine experimentally.

It’s been a truth for a long time: if you want to study the movement and behavior of single atoms, electron microscopy can give you what X-rays can’t. X-rays are good at penetrating into samples—they allow you to see what happens inside batteries as they charge and discharge, for example—but historically they have not been able to spatially image with the same precision electrons can.

But scientists are working to improve the image resolution of X-ray techniques. One such method is X-ray tomography, which enables non-invasive imaging of the inside of materials. If you want to map the intricacies of a microcircuit, for example, or trace the neurons in a brain without destroying the material you are looking at, you need X-ray tomography, and the better the resolution, the smaller the phenomena you can trace with the X-ray beam.

To that end, a group of scientists led by the U.S. Department of Energy’s (DOE) Argonne National Laboratory has created a new method for improving the resolution of hard X-ray nanotomography. (Nanotomography is X-ray imaging on the scale of nanometers. For comparison, an average human hair is 100,000 nanometers wide.) The team constructed a high-resolution X-ray microscope using the powerful X-ray beams of the Advanced Photon Source (APS) and created new computer algorithms to compensate for issues encountered at tiny scales. Using this method, the team achieved a resolution below 10 nanometers.

New research has found that artificial intelligence (AI) analyzing medical scans can identify the race of patients with an astonishing degree of accuracy, while their human counterparts cannot. With the Food and Drug Administration (FDA) approving more algorithms for medical use, the researchers are concerned that AI could end up perpetuating racial biases. They are especially concerned that they could not figure out precisely how the machine-learning models were able to identify race, even from heavily corrupted and low-resolution images.

In the study, published on pre-print service Arxiv, an international team of doctors investigated how deep learning models can detect race from medical images. Using private and public chest scans and self-reported data on race and ethnicity, they first assessed how accurate the algorithms were, before investigating the mechanism.

“We hypothesized that if the model was able to identify a patient’s race, this would suggest the models had implicitly learned to recognize racial information despite not being directly trained for that task,” the team wrote in their research.