The constant struggle between viruses and our immune system is similar to the way we interpret words. Researchers at MIT have applied machine learning tools to identify protein regions that can help coronavirus and other pathogens escape antibodies and vaccines.
In 1950, Alan Turing, one of the fathers of computers, predicted that machines would compete with people in the “field of knowledge” and argued that they could even learn to understand and speak English. This was a very ambitious The goal is because although regular grammars help sentence construction, they are difficult to infer meaning.
inside Human natural language There are many ways to express the same idea, and words used in the same context often have similar meanings. However, small changes in letters can completely change the meaning of a sentence.
To meet these challenges and train computers, scientists have developed language-based natural language processing tools. Machine learning, Now, from Massachusetts Institute of Technology, In the United States) was inspired by them and applied them to a completely different field: understanding how viruses evade our body’s defenses.
Based on the way we use words, the researchers published a paper in the journal this week. science A new approach Identify and predict mutations (Variations in the amino acid sequence of the protein) allow the virus Escape Human immunity and vaccines. In this way, expensive experimental techniques currently used for the same purpose can be avoided or reduced.
Consequences of exchange of letters
One of the authors Bryan Bryson, Provides a language example for SINC: “Let’s consider the phrase in English Boy pOnedog (The boy pats the dog). Just change one letter and we can continue to retain the syntax and semantics: Boy pEdog (Boy petting dog), but also lost grammar correction function: Boy patX That dog (Patx does not exist)”.
“However, if only by changing one character, we want it to follow the English rules, thereby substantially changing the meaning, then we can say: boy EArtus (The boy ate the dog) “It has nothing to do with the previous sentence.
Similarly, the authors found that viruses can evade immune responses through mutations. These mutations retain the biological “grammar or syntax” that controls the infectivity of the virus, but change the “semantics” or meaning of the protein sequence, so it will not be recognized by antibodies and infected cells. .
This ability of the virus represents a major new challenge for human development. Vaccines and antiviral drugs, Especially in the manufacture of widespread flu and effective HIV therapy. In the covid-19 pandemic, this is’Escape the virusWhen looking for a solution to the coronavirus, it has also become a priority.
He said: “Using public data (original virus original sequence), we optimize the model’s high semantic conversion while maintaining high syntax (especially for influenza viruses), and can identify “rich” mutations for the virus to escape. “He said.Bryson
“He went on to say that what we show in the article is that we can find A more or less escaped area or domain. For example, we showed that the “head” of the influenza virus hemagglutinin (HA) protein is more likely to do so than the “stem”, which is consistent with the results seen by influenza vaccine researchers after many exercises.
In addition to influenza virus proteins, the results of this model can also accurately predict mutations and regions related to virus immune escape. VIH virus The coronavirus that causes AIDS and the covid-19 pandemic.
“in order to SARS-CoV-2 Spike Protein, Our model predicts that the two protein domains (receptor binding and the so-called N-terminus) are more likely to escape than the other region of the protein called S2 “Bryson interpretation”, and we can use this information to design other Experiment in the laboratory and explore which protein regions the therapeutic antibody or vaccine produced antibody binds to.”
“The importance of all of this is that when you design a new antiviral drug or develop a vaccine, you may wish to An area where the target is unlikely to escapeThe MIT researchers concluded that these areas will become more stable over time.
Brian Hie, Ellen D. Zhong, Bonnie Berger and Bryan Bryson: “Learn the language of virus evolution and escape.” (Angle: Y.-A. Kim y TM Przytycka. “The language of viruses”). science, 2021