Filippo Costanti (University of Siena)

When: Nov 24th, 2021 – 11:45 – 12:30 AM

Where: Google meet link

Description

Recent research studies have highlighted that individuals not exposed to SARS coronavirus 2 (SARS-CoV2) have pre-existing immune system reactivity, due to contact with human coronaviruses that cause the common cold. Therefore, a more in-depth comparative study of the proteins of the different human coronaviruses becomes fundamental to face the new COVID19 pandemic. The SARS-CoV2 genome encodes structural and non-structural proteins, some of which have shown significant cross-activation of cells of the immune system. The comparison of these proteins, through alignment techniques of amino acid sequences, for example based on algorithms such as BLAST, can provide a basis for future biological studies especially for the production of a vaccine that, in addition to the well-known spike protein (S), uses other proteins as targets.
Another very important aspect from a bioinformatics point of view is the ability to manage the increasing amount of data derived from in vitro analysis and sequencing techniques, in order to compare nucleotide and amino acid sequences with greater effectiveness compared to classical algorithms. The use of artificial intelligence (AI) in this area has proved decisive and has produced surprising results, for example, in modeling the 3D structure of proteins for the discovery of new drugs.
In this thesis, coronavirus proteins  historical strains and SARS-CoV-2  were comparatively analyzed to highlight their homologies, using both classical alignment methods and Siamese neural networks. In particular, the Siamese networks, trained on homologous proteins at different levels of similarity, are able to calculate, in an extremely efficient way, an alignment score comparable to that produced by BLAST.