The Conversation
Illustration: IEEE Spectrum; Model: Getty Images
In March, organizers of the COVID Moonshot initative crowdsourced chemical designs for COVID-19 antivirals. They received over 14,000 submissions from chemists around the world.
<p><a href="https://postera.ai">PostEra</a>, a machine-learning company leading the <a href="/the-human-os/artificial-intelligence/medical-ai/covid-moonshot-can-ai-algorithms-and-volunteer-chemists-design-a-knockout-antiviral">Moonshot initiative</a>, triaged those submissions for how quickly and easily each chemical compound could be synthesized. One looked particularly promising, and PostEra sent data about the compound back to an online volunteer crowd of medicinal chemists.</p>
<p></p>
<p>The crowd and PostEra’s machine-learning algorithms iterated back and forth, designing and testing tweaks on the chemical structure. Soon, the compound’s potency had increased by two orders of magnitude. Then, the chemical compound successfully killed live coronavirus in human cells without harming the cells. Now, that drug candidate and three more promising compounds are headed to animal testing in preparation for human clinical trials.</p>
<!--nextpage-->
<p></p>
<p>“All four are progressing really rapidly,” says <a href="https://www.alpha-lee.com/">Alpha Lee</a>, cofounder and chief scientific officer of PostEra. “Now we are moving to the next phase: We are launching the drugs in animals and paving a path to the clinic.”</p>
<p></p>
<p>The COVID Moonshot, an open-science initiative that <a href="/the-human-os/artificial-intelligence/medical-ai/covid-moonshot-can-ai-algorithms-and-volunteer-chemists-design-a-knockout-antiviral">combines crowdsourcing with high throughput crystallography and machine learning</a>, has synthesized and tested 1,000 compounds in less than 6 months, including generating crystal structures for over 200 of the compounds. Over 30 teams and organizations, including large university labs, chemical synthesis companies, and pharmaceutical companies, have provided time, expertise, and materials pro-bono or at cost.</p>
<p></p>
<img type="lazy-image" data-runner-src="https://assets.rebelmouse.io/eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpbWFnZSI6Imh0dHBzOi8vYXNzZXRzLnJibC5tcy8yNTUzNjQzMi9vcmlnaW4uanBnIiwiZXhwaXJlc19hdCI6MTYyNDM0MzY4Nn0.qnHpl98YrKIP0bAT2azl3J0OXSh4kUK3fkcfAwKlSy0/img.jpg?width=980" id="0500a" class="rm-shortcode" data-rm-shortcode-id="e1cc30769799268a4c06f1972abd60f5" data-rm-shortcode-name="rebelmouse-image" alt="Global map of contributions to the COVID Moonshot project" />
<p>Currently, the group has narrowed in on four chemical series—families of structurally-related compounds—that each show antiviral activity and drug-like qualities such as being stable in the body and potent in small amounts. “Four starting points evolved and matured into four elites,” says Lee. Having four chemically-diverse options means that if one turns out to be lacking, the group can quickly switch to another, he adds.</p>
<p></p>
<p>The compounds target a protein that is the Achilles heel for coronaviruses, the coronavirus main protease, <a href="https://www.rcsb.org/structure/1P9S">Mpro</a>. Which means the drugs could be effective against all coronaviruses, not just SARS-CoV-2. Additionally, small molecule drugs are often easier to make and distribute than vaccines, so Lee hopes new antivirals could help combat the current pandemic in places where vaccines may be hard to access.</p>
<p></p>
<p>The Moonshot team maintains a living summary of their <a href="https://www.biorxiv.org/content/10.1101/2020.10.29.339317v1">data and results</a> on bioRxiv, and everything produced by the group—including all data and final drug designs—is being made <a href="https://discuss.postera.ai/t/covid-project-faq-about-us/72">openly available</a> with no intellectual property restrictions. “You can think of this as a generic drug from day zero,” says Lee.</p>
<p></p>
<img type="lazy-image" data-runner-src="https://assets.rebelmouse.io/eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpbWFnZSI6Imh0dHBzOi8vYXNzZXRzLnJibC5tcy8yNTUzNjQzMy9vcmlnaW4uanBnIiwiZXhwaXJlc19hdCI6MTYzMzI0MTA2NX0.24o-Xw8zoFIp9Rl3j1zDnz6YozT1dNhs8HpN26ZDzgc/img.jpg?width=980" id="7a8b9" class="rm-shortcode" data-rm-shortcode-id="e020d70b105138554815169446a2756e" data-rm-shortcode-name="rebelmouse-image" alt="Iterations on a crowdsourced drug design against coronavirus led to a 170x improvement in potency." />
<p>To fund animal studies—a series of preclinical tests that measure safety parameters and optimize drug-like properties of a compound—the non-profit initiative has begun a $1.5 million <a href="https://helpcurecovid.org/">fundraising campaign</a>. Once the animal studies are complete in mid-2021, Lee hopes pharmaceutical collaborators will be eager to put the antiviral into clinical development, even without intellectual property rights, because all the expensive drug discovery and preclinical tests are already complete.</p>
<p></p>
<p>In June, Lee and three other leaders of the Moonshot <a href="https://www.nature.com/articles/s41557-020-0496-2">published an editorial</a> on the benefits of crowdsourcing drug design. The COVID Moonshot wants to lead by example, proving that a crowdsourced drug discovery process can successfully lead to human clinical trials. “I think one of the lasting impacts of Moonshot is not only a pan-coronavirus drug to cure COVID and prevent future pandemics, but also to nucleate a new way of organizing drug discovery,” says Lee.</p>
<div class="mobileShow">
<h3 class="RptHdBackBarMobile"><span class="BackArrowBlkBkgrd"><</span> <a href="/static/covid19-ieee-resources">Back to IEEE COVID-19 Resources</a></h3>
</div>
Keep reading...
Show less
Illustration: IEEE Spectrum
The death penalty, abortion, gun legislation: There’s no shortage of controversial topics that are hotly debated today on social media. These topics are so important to us because they touch on an essential underlying force that makes us human, our morality.
Researchers in Brazil have developed and analyzed three models that can describe the morality of individuals based on the language they use. The results were published last month in IEEE Transactions on Affective Computing.
<p>Ivandré Paraboni is an associate professor at the <span>School of Arts, Sciences and Humanities </span>at the University of São Paulo who led the study. His team choose to focus on a theory commonly used by social scientists called <a href="https://en.wikipedia.org/wiki/Moral_foundations_theory">Moral foundations theory</a>. It postulates several key categories of morality including care, fairness, loyalty, authority, and purity. </p>
<p>The aim of the new models, according to Paraboni, is to infer values of those five moral foundations just by looking at their writing, regardless of what they are talking about. “They may be talking about their everyday life, or about whatever they talk about on social media,” Paraboni says. “And we may still find underlying patterns that are revealing of their five moral foundations.”</p>
<p>To develop and validate the models, Paraboni’s team provided more than 500 volunteers with questionnaires. Participants were asked to rate eight topics (<em>e.g</em>., same sex marriage, gun ownership, drug policy) with sentiment scores (from 0 = ‘totally against’ to 5 = ‘totally in favor’). They were also asked to write out explanations of their ratings.</p>
<p>Human judges then gave their own rating to a subset of explanations from participants. The exercise determined how well humans could infer the intended opinions from the text. “Knowing the complexity of the task from a human perspective in this way gave us a more realistic view of what the computational models can or cannot do with this particular dataset,” says Paraboni.</p>
<p>Using the text opinions from the study participants, the research team created three machine learning algorithms that could assess the language used in each participant’s statement. The models analyzed psycholinguistics (emotional context of words), words, and word sequences, respectively.</p>
<p>All three models were able to infer <span>an individual's moral foundations from the text. T</span>he first two models, which focus on individual words used by the author, were more accurate than the deep learning approach that analyzes word <span>sequences</span>.</p>
<p>Paraboni adds, “Word counts–such as how often an individual uses words like ‘sin’ or ‘duty’–turned out to be highly revealing of their moral foundations, that is, predicting with higher accuracy their degrees of care, fairness, loyalty, authority, and purity.”</p>
<p></p>
<p>He says his team plans to continue to incorporate other forms of linguistic analysis into their models. They are, he says, exploring other models that focus more on the text (independent of the author) as a way to analyze Twitter data.</p>
Keep reading...
Show less
Load more