It’s 2021, and the world has a safe vaccine for COVID-19 as well as drugs to treat the disease. But now, we face a serious problem: How do we make billions of doses of the medicines?

Researchers fear that the same issues with supply chains that caused toilet paper shortages at the beginning of the pandemic in the United States may result in the same problem with the fine chemicals needed to synthesize COVID-19 therapeutics and vaccines.

Now, a University of Michigan team of medicinal chemists have used artificial intelligence to find alternative pharmaceutical building blocks for 12 drugs under investigation to treat COVID-19.

“The WHO has started to discuss who will be the first to receive vaccinations for COVID-19, should they become available,” said U-M researcher Tim Cernak, an assistant professor of medicinal chemistry and chemistry.

“Just to hit that front group—the essential workers, the sick, the elderly—we will need 4.2 billion doses of a vaccine, because the dosing regimen is at least two doses per person and there will be wastage. For us synthetic chemists, the folks who actually produce the medicine, that number is mind boggling. We’re in a supply chain crisis.”

For example, the researchers say the only approved drug for treating COVID-19 is remdesivir. The drug, an antiviral nucleotide, had an availability of just 5,000 when the outbreak started. As of the week of July 13, the United States is adding approximately 60,000 cases of coronavirus infection a day.

Cernak and his lab were approached by chemical supplier MilliporeSigma to devise solutions to the supply issue. Cernak and his team combed the federal clinical trials database for drugs currently being considered for treatment of COVID-19, and then used the artificial intelligence software Synthia to determine new ways to piece the drugs

Patient data are a treasure trove for AI researchers. There’s a problem though: many algorithms used to mine patient data act as black boxes, which makes their predictions often hard to interpret for doctors. Researchers from Eindhoven University of Technology (TU/e) and the Zhejiang university in China have now developed an algorithm that not only predicts hospital readmissions of heart failure patients, but also tells you why these occur. The work has been published in BMC Medical Informatics and Decision Making.

Doctors are increasingly using data from electronic healthcare records to asses patient risks, predict outcomes, and recommend and evaluate treatments. Application of machine learning algorithms in clinical settings has however been hampered by lack of interpretability. The models often act as black boxes: you see what goes in (data) and what comes out (predictions), but you can’t see what happens in between. It can therefore be very hard to interpret why the models are saying what they are saying.

This undermines the trust healthcare professionals have in machine learning algorithms, and limits their use in everyday clinical decisions. Of course, interpretability is also a key requirement of EU privacy regulations (GDPR), so improving it also has legal benefits.

Attention-based neural networks

To solve this problem, Ph.D. candidate Peipei Chen of the Department of Industrial Engineering and Innovation Sciences, together with other researchers at TU/e and Zhejiang University in Hangzhou, has tested an attention-based neural network on heart patients in China. Attention-based networks are able to focus on key details in data using contextual information.

The Korea Brain Research Institute (headed by Suh Pann-ghill) announced on July 24 the discovery that an increase in amyloid-beta in the brain alters cholesterol biosynthesis, which was found by Dr. Cheon Mookyung of KBRI through RNA-seq analysis data (omics) using AI.

The research results were published in PLOS Computational Biology, an international academic journal in the field of computational biology.

Amyloid-beta is known as a protein that causes Alzheimer’s disease (AD). In a normal brain, if it accumulates excessively, it is removed by microglia, etc. Cholesterol must also be kept at a certain level within the blood to compose the cell membrane, adjust fluidity in the membrane, and maintain homeostasis. If the abovementioned processes are not carried out properly, pathological abnormalities occur in the body.

The research team analyzed the cerebral cortex tissue data of AD mouse models using the cutting-edge deep learning technique of generative adversarial networks (GANs). A GAN is an algorithm that generates data through competition between a generator and discriminator, studies the generated data, and creates synthetic data close to real images. The GANs technique was used to create the fake speech video of President Barack Obama and can be applied to continuous face aging.

The research team performed an AD genetic expression simulation in mice using GANs and observed the changing process of gene expression from normal to AD state. As a result, it was discovered that amyloid-beta increases and alters cholesterol biosynthesis in the early stages of the disease. This discovery was also confirmed by the RNA-seq analysis of postmortem brain tissue.

This means that the increase in amyloid-beta triggers cholesterol biosynthesis and that the two processes in combination are likely to