NVIDIA model can predict coronavirus variants

NVIDIA’s GenSLMs language model was able to use data from the SARS-CoV-2 virus to predict mutations. The technology was trained with the initial genomic sequence and, using high-performance computing and AI, generated subvariants, two of them corresponding to the Eris and Pirola variants, which prevailed in the world this year.

Read more:

NVIDIA Language Model and AI

• GenSLMs is a model fed with genomic data from viruses and one of its fundamental features is that it can interpret long sequences of nucleotides (subunits that form DNA and RNA) just as a normal AI model would read a sentence.
• This allows him to understand the relationship between the virus that causes Covid-19 and its possible variations, which generated the variant predictions.
• Among the sequences of possible variants generated by AI, two have already been seen in real life: Eris and Pirola.
• It is also worth highlighting that the AI ​​was trained with data from 2020, when the variants had not yet appeared, proving the effectiveness of the model in predicting mutations.
• The project’s lead researcher, Arvind Ramanathan, stated in a note how the model is “extremely naive, without any specific information or restrictions on what a new COVID variant would look like.” In other words, she came up with that variation on her own.

Demonstration

In addition to generating variations, NVIDIA’s language model can also classify coronavirus genome sequences by distinguishing between identified variants.

A demonstration will take place in November at NGC, NVIDIA’s accelerated software center, and will allow users to explore visualizations of GenSLMs analysis, including Covid. They will be able to choose from eight variants to understand how AI tracks mutations, indicating which components will remain and which will change in a mutation.

How AI can help fight Covid-19

According to Ramanathan, AI helps understand how the Covid-19 virus evolves, which helps understand how a specific strain can escape the immune system or even vaccines. This could allow health systems and vaccine manufacturers to prepare.

Understanding how different parts of the genome are co-evolving gives us clues about how the virus might develop new vulnerabilities or new forms of resistance.

Arvind Ramanathan