Science Paper Interpretation: Techniques such as viral surface protein modeling help develop universal vaccines

In a new study, researchers from the Massachusetts Institute of Technology in the United States have now designed a new method based on models originally developed to analyze languages to compute virus escape models.
the model can predict which parts of the virus's surface protein are more likely to mutate, allowing the virus to escape, and it can also identify those that are unlikely to mutate, making it a good target for developing new vaccines.
study was published in the January 15, 2021 issue of the journal Science under the title "Learning the language of viral evolution and escape."
virus escape is a big problem," said Bonnie Berger, head of the computational and biology team at MIT's Computer Science and Artificial Intelligence Laboratory.
escape of the surface protein of the influenza virus and the surface protein of HIV are the causes of the fact that we do not have a universal influenza vaccine and an HIV vaccine, both of which kill hundreds of thousands of people each year.
study, Berger and her colleagues identified potential targets for developing vaccines against influenza viruses, HIV and SARS-CoV-2.
the paper was accepted for publication, the researchers also applied their models to the recent emergence of new SARS-CoV-2 variants in the UK and South Africa.
said the yet-to-be peer-reviewed analysis suggests that the genetic sequence of the virus variant should be further investigated to determine whether they are likely to escape the effects of existing vaccines.
Berger and Bryan Bryson, assistant professor of bioengineering at the Massachusetts Institute of Technology, are the authors of the paper.
first author of the paper is Brian Hie, a graduate student at the Massachusetts Institute of Technology.
different types of viruses get genetic mutations, HIV and influenza viruses are among the fastest-mutated viruses.
these mutations to facilitate virus escape, they must help the virus change the shape of its surface protein so that antibodies can no longer bind to it.
, however, changes in proteins do not de-functionalize them.
the researchers decided to model these standards using a computational model called a language model from the field of natural language processing (NLP).
models were originally designed to analyze patterns in a language, especially the frequency with which certain words appear.
, these models can predict which words can be used to complete a sentence, such as "Sally eats eggs for..."
the selected word must be both grammatically correct and correct in meaning.
example, the NLP model might predict "breakfast" or "lunch."
insights of these researchers is that this type of model can also be applied to biological information, such as gene sequences.
, syntax is similar to the rules that determine whether a protein encoded in a particular sequence has function, while semantics are similar to whether a protein can take a new shape to help it evade antibodies.
, mutations that allow viruses to escape must maintain the syntax of the sequence, but change the structure of the protein in a useful way.
wants to escape the human immune system and it doesn't want to mutate itself, it will die or not replicate," he said.
wants to be adaptable, but pretend it's good enough so that it can't be detected by the human immune system.
to model this process, the researchers trained an NLP model to analyze patterns found in gene sequences, which allowed them to predict new sequences with new functions that still follow biological rules of protein structure.
advantage of this modeling is that it requires only sequence information, which is easier to obtain than protein structures.
the model could be trained in a relatively small amount of information--- in the study, they used 60,000 HIV sequences, 45,000 influenza virus sequences and 4,000 coronavirus sequences.
is very powerful because it learns about this complex distribution structure and only learns some functional insights from sequence changes, " says Hie.
we have this large library of the virus sequence data for each amino acid location, and this model can learn the characteristics of amino acid co-occurrence and co-mutation in the training data.
" blocking virus escape Once the model was trained, the researchers used it to predict sequences of coronavirus tingling proteins, HIV envelope proteins, and influenza virus hemoglobin (HA) proteins that more or less produce escape mutations.
for influenza viruses, the model shows that the most unlikely sequence of mutations and virus escapes is the handle of the HA protein.
this is consistent with recent research --- antibodies to the HA protein shanks, which are not produced by most people infected with or vaccinated against influenza--- providing near-universal protection against any influenza strain.
of coronavirus from this model shows that a portion of the prickly protein called S2 sub-base is the least likely to produce escape mutations.
how quickly the SARS-CoV-2 virus mutates remains a question, it remains to be known how long the currently deployed vaccine against the COVID-19 pandemic will last.
evidence suggests that the virus does not mutate as fast as the flu virus or HIV.
, however, scientists have recently discovered new mutations in Singapore, South Africa and Malaysia that they believe should be investigated for potential virus escapes (the new data have not yet been peer-reviewed).
in their study of HIV, the researchers found that there were many possible escape mutations in the V1-V2 hyperconverter region of the envelope protein, consistent with previous studies, and found sequences with a lower probability of escape.
researchers are working with others to identify potential targets for cancer vaccines that stimulate the body's own immune system to destroy tumors.
, they say, could also be used to design small molecule drugs that may be less susceptible to drug resistance to diseases such as tuberculosis.
, "There are a lot of opportunities, and the best thing is that what we need is sequence data that's easy to produce."
" (Bio Valley Bioon.com) Reference: 1.Brian Hie el al. Learning the language of viral evolution and escape. Science, 2021, doi:10.1126/science.abd7331. 2.Yoo-Ah Kim el al. The language of a virus. Science, 2021, doi:10.1126/science.abf6894. 3.Model analyzes how viruses escape the immune system