Manufacturing Bits: May 11

Covid-19 data mining
Using machine learning and other technologies, Lawrence Berkeley National Laboratory (Berkeley Lab) has developed a data text-mining tool to help synthesize a growing amount of scientific literature on Covid-19.

Each day, some 200 new journal articles are being published on the coronavirus alone, according to Berkeley Lab. Berkeley Lab’s data mining tool, which is live on covidscholar.org, uses natural language processing techniques that scan and search thousands of research papers. It helps draw insights and connections to the data.

The tool runs on supercomputers from the National Energy Research Scientific Computing Center (NERSC), a DOE Office of Science user facility located at Berkeley Lab. The online search engine and portal are powered by the Spin cloud platform at the NERSC.

The hope is that the tool could eventually enable “automated science.” Using the tool after a month, Berkeley Lab has collected over 61,000 research papers. About 8,000 of them are about Covid-19, while the rest are about related topics like viruses and pandemics.

“On Google and other search engines people search for what they think is relevant,” said Berkeley Lab scientist Gerbrand Ceder. “Our objective is to do information extraction so that people can find nonobvious information and relationships. That’s the whole idea of machine learning and natural language processing that will be applied on these datasets.”

AI antibody searches
With machine learning, Lawrence Livermore National Laboratory (LLNL) has identified a set of therapeutic antibody sequences that are targeted at binding and neutralizing SARS-CoV-2, the virus that causes Covid-19.

Antibodies are harvested from the blood of patients who have fully recovered from Covid-19. Antibodies are potential treatments for the virus. LLNL’s goal is to develop new and improved antibody designs using iterative computational-experimental processes.

For this, LLNL used high-performance computers and machine learning to find antibody candidates that bind with the SARS-CoV-2 Receptor Binding Domain (RBD). Researchers used more than 200,000 CPU hours and 20,000 GPU hours on two high-performance computers at LLNL–Corona and Catalyst. The system has performed nearly 180,000 calculations of candidate antibodies with the SARS-CoV-2 RBD.

Using these technologies, researchers have narrowed down the number of possible designs from a nearly infinite set of candidates to 20 initial sequences predicted to target SARS-CoV-2.

“Our computational results are encouraging, and we’re excited about the experimental tests that are underway now,” said LLNL data scientist Dan Faissol. “We hope that one of these initial antibody designs binds to the SARS-CoV-2 target as intended, but regardless of the outcome, the experimental results will significantly improve our ability to design a subsequent round of antibodies.”

Virus modeling
Using a supercomputer and modeling techniques, Johannes Gutenberg University Mainz (JGU) has identified several potential active substances to combat coronavirus.

Several drugs approved for treating the hepatitis C viral infection were identified as potential candidates against Covid-19, according to JGU.

This is the result of research based on calculations using the MOGON II supercomputer at JGU. JGU operate a supercomputer that provides more than 53,000 CPU cores with a peak power of 2 petaFLOPS ( 2 trillion floating-point operations per second).

Researchers used a technique called molecular docking. “The molecular docking approach can be used to model the interaction between a small molecule and a protein at the atomic level, which allow us to characterize the behavior of small molecules in the binding site of target proteins as well as to elucidate fundamental biochemical processes,” according to Jilin University and others in a paper.

Using the MOGON II, researchers simulated the way that about 42,000 different substances listed in open databases bind to certain proteins of SARS-CoV-2. This in turn could potentially inhibit the penetration of the virus into the human body or its multiplication.

With the system, researchers made more than 30 billion single calculations within two months. They found that compounds from the four hepatitis C drugs–simeprevir, paritaprevir, grazoprevir, and velpatasvir–have a high affinity to bind SARS-CoV-2 strongly and may therefore be able to prevent infection.

“This computer simulation method is known as molecular docking and it has been recognized and used for years. It is much faster and less expensive than lab experiments,” said Thomas Efferth, a professor of the JGU Institute of Pharmacy and Biomedical Sciences. “As far as we know, we were the first to have used molecular docking with SARS-CoV-2. And it is fantastic news that we have found a number of approved hepatitis C drugs as promising candidates for treatment.

“This is also supported by the fact that both SARS-CoV-2 and the hepatitis C virus are a virus of the same type, a so-called single-stranded RNA virus,” explained Efferth. “Our research results now need to be checked in laboratory experiments and clinical studies.”