Predicting the upcoming variant

When initial reports of a novel COVID-19 variant surface, researchers globally rush to tackle an essential question: Will this new strain spread more easily or display greater severity than its forerunners? Often, by the time a response is determined, it is too late to influence immediate public health policies or modify vaccination strategies, resulting in a loss of crucial time, effort, and resources for health officials.

In two recent publications in Proceedings of the National Academy of Sciences (PNAS), a research group in the Department of Chemistry and Chemical Biology meshed biophysics with artificial intelligence to swiftly identify high-risk viral variants — presenting a groundbreaking method for managing pandemics. Their objective is to anticipate a virus’s evolutionary shifts before it poses a risk to public health.

“As a society, we frequently find ourselves quite unprepared for the emergence of new viruses and pandemics; thus, our laboratory has been focused on methods to be more anticipatory,” stated senior author Eugene Shakhnovich, Roy G. Gordon Professor of Chemistry. “Utilizing fundamental concepts from physics and chemistry, we crafted a multiscale model to predict the evolutionary trajectory of a specific variant and to foresee which variants will ascend in populations.”

The analyses outline strategies for predicting the viral variants most likely to escalate into public health threats while also accelerating experimental validation. Collectively, these advancements redefine both the forecasting and identification of perilous viral variants, establishing a framework for more extensive applications.

These investigations were spearheaded by members of Shakhnovich’s laboratory, including co-authors Dianzhuo (John) Wang and Vaibhav Mohanty, both Ph.D. students at the Harvard Kenneth C. Griffin Graduate School of Arts and Sciences, along with Marian Huot, a visiting student from École Normale Supérieure.

“This framework not only assists us in monitoring variants — it also enables us to stay ahead of them.”

Marian Huot, visiting student and co-author

“Our research has concentrated on the spike protein of COVID-19, investigating how its mutations affect viral adaptability and immune evasion,” Wang explained. “Considering that COVID-19 is the most comprehensively documented pandemic thus far, we recognized an opportunity to create models that not only comprehend viral evolution but also predict which mutations are likely to present the most significant challenges.”

The initial study presented a model that quantitatively related biophysical characteristics — such as the binding affinity of the spike protein to human receptors and its capacity to avoid antibodies — to a variant’s potential for proliferation within global populations. By integrating a complex, yet vital element called epistasis (where the effect of one mutation is dependent on another), the model overcame a major limitation seen in previous methodologies that had difficulty in rendering accurate predictions.

“Evolution isn’t straightforward — mutations interact, sometimes unlocking new routes for adaptation,” Shakhnovich remarked. “Incorporating these relationships allowed us to predict the appearance of dominant variants ahead of epidemiological indicators.”

Building on these observations, the accompanying study unveils VIRAL (Viral Identification via Rapid Active Learning), a computational structure that blends the biophysical model with artificial intelligence to expedite the identification of high-risk SARS-CoV-2 variants. By examining potential mutations in the spike protein, it discerningly pinpointed those most likely to heighten transmissibility and immune evasion.

“At the onset of a pandemic, when experimental resources are limited, we cannot afford to evaluate every possible mutation,” Wang stated. “VIRAL employs artificial intelligence to concentrate lab efforts on the most pressing candidates — significantly accelerating our capacity to recognize variants that could instigate the next wave.”

The ramifications of this research are extensive. Simulations demonstrate that the VIRAL framework can pinpoint high-risk SARS-CoV-2 variants up to five times more swiftly than traditional strategies, while requiring less than 1 percent of experimental screening efforts. This pronounced increase in efficiency could substantially enhance early outbreak responses.

“This framework not only assists us in monitoring variants — it allows us to stay ahead of them,” said Huot. “By detecting high-fitness variants before they manifest in the general population, we can devise vaccination strategies that proactively address, rather than merely react to, emerging threats.”

A standout characteristic of this work is its interdisciplinary nature, with the international Harvard team integrating fields of molecular biophysics, artificial intelligence, and virology to deepen our comprehension of swiftly evolving viral threats.

“By merging physics-driven modeling with machine learning, we are establishing a predictive framework for viral evolution with vast potential,” Shakhnovich noted. “We are eager to explore how this approach may extend beyond infectious diseases into realms like cancer biology.”

Looking forward, the team plans to modify and expand the framework for broader applications, addressing challenges such as other emerging viruses and rapidly evolving tumor cells. They stress that the convergence of physical modeling with AI may shift the paradigm from reactive monitoring to proactive biological forecasting.

“In a world where biological challenges are in constant flux, early warnings and intelligent tools are crucial,” Wang emphasized. “Our ultimate ambition is to develop a platform — one that equips scientists and policymakers with a head start not only in impending pandemics but in confronting swiftly evolving challenges across biological contexts,” added Huot.

Shakhnovich acknowledged funding from the National Institutes of Health for enabling exploratory research aimed at benefitting public health. Essential scientific inquiry and future innovations face severe threats due to budget cuts in Washington, Shakhnovich cautioned.

“Our research has the capability to assist all of humanity in addressing notable health issues,” Shakhnovich mentioned. “Such progress would not have been achievable without federal backing that seeks long-term advantages.”