Fine-tuning innovation to meet COVID challenges — GCN

Fine-tuning innovation to meet COVID challenges

The [email protected] distributed computing project has reached exascale processing power with the quickly growing number of citizen scientists running simulations of protein dynamics on their PCs to find protein structures that drugs could target to fight and treat COVID-19. 

Computationally expensive molecular science is usually conducted on supercomputers, but the program’s expanding user base has generated an exaflop of computing power, or a billion billion calculations per second, making it 10 times faster than Summit, the world’s fastest publicly ranked 200- supercomputer at Oak Ridge National Lab.

Summit has been running molecular dynamics simulations on the virus’ main “spike” protein, hoping to found a compound that would bind to the spike, interfering with the infection process. Scientists at Oak Ridge were able to identify 77 candidate compounds (such as medications) that warranted further experimentation, helping to narrow the field for medical researchers.

However, researchers needed more analysis of structure and behavior of COVID-19’s viral proteins, so [email protected]’s Director Greg Bowman committed the program to developing an antibody that could help to prevent the virus from infecting lung cells, like the one developed for SARS-CoV.

A rush of volunteers even caused some downtime as the [email protected] team rushed to set up more simulations for volunteers to run. 

VMware developed an appliance for [email protected] that runs in vSphere environments along with a vRealize operations dashboard to query the [email protected] APIs and report statistics. 

AI assist

Researchers at the University of Massachusetts Amherst have invented a portable surveillance device that listens for coughing and sneezing and analyzes the audio data to detect flu-like illnesses and influenza trends.

Powered by machine learning, the FluSense edge-computing platform was envisioned to be used in in hospitals, health care waiting rooms and larger public spaces.

It uses a Raspberry Pi and neural computing engine for analyzing data from a low-cost microphone array and a thermal camera, storing no personally identifiable information, such as speech data or distinguishing images.

The researchers first developed a lab-based cough model and trained the deep neural network classifier to draw boxes on thermal images representing people and then to count the coughing individuals. They tested the dictionary-sized FluSense devices in four waiting rooms at the university’s health clinic from December 2018 to July 2019.

In that time, the FluSense platform collected and analyzed more than 350,000 thermal images and 21 million non-speech audio samples from the public waiting areas.

The researchers found that FluSense was able to accurately predict daily illness rates at the university clinic. Multiple and complementary sets of FluSense signals “strongly correlated” with laboratory-based testing for flu-like illnesses and influenza itself.

According to the study, “the early symptom-related information captured by FluSense could provide valuable additional and complementary information to current influenza prediction efforts,” such as the FluSight Network, which is a multidisciplinary consortium of flu forecasting teams.

In the private sector, Austin-based Athena Security is developing an COVID19 screening component as part of its AI-enabled gun detection security camera platform that it says will be able to detect fevers in people and alert camera owners that people may be carrying the coronavirus.

“Since higher temperature is one of the first symptoms, these cameras can be life-saving— warning the person that they could have the virus and encouraging that person to take serious steps to self-quarantine,” the company  told Motherboard.

Frictionless finance

An early version of the recently-passed coronavirus relief package included a plan for a government-run digital payment platform to disperse federal funds.

The early draft of the House Democrats’ proposed stimulus package called on the Federal Reserve to make ‘digital dollar wallets’ available to U.S. citizens, legal permanent residents and businesses, where stimulus funds could be deposited directly from the Federal Reserve, according to a report in MIT Technology Review.

Too big a lift on too short notice to make it into the final bill, the idea behind FedAccounts is that citizens and business could have accounts with the Federal Reserve where digital currency – much like what the Fed already uses with banks – could be deposited. The accounts would offer “high interest, instant payments, and full government backing [and] … all the functionality of ordinary bank accounts with the exception of overdraft coverage,” according to a 2018 paper on the concept.

“The FedAccount program would put government-issued digital or ‘account’ money on par with government-issued physical currency, transforming digital dollars into a resource that anyone can use,“ the authors wrote. The system would seamlessly interoperate with the existing and reliable technology the Fed currently uses for money and payments. 

The Senate also floated a version of the idea. The Banking for All Act, introduced March 24 by Sen. Sherrod Brown, (D-Ohio), would require banks to maintain “digital dollar wallets,” which would let the millions of unbanked consumers access coronavirus stimulus payments by setting up FedAccounts at local banks and post offices. Account holders would receive debit cards, online account access, automatic bill pay, mobile banking and ATM access at post offices, Brown said in his announcement.