Spyros Lytras

This study analyzes poxvirus gene repertoires and finds that they evolve through expansion-biased birth-and-death dynamics and genome architectural constraints.

This study identifies key ACE2 residues that influence sarbecovirus infectivity in African Rhinolophus bats and reports novel ACE2 sequences from R. simulator and R. blasii.

This study evaluates how protein language models capture mutational processes and fitness landscapes, finding that explicit mutational accessibility improves predictions while historical biases degrade performance.

The paper develops a host-response-based framework to detect viral infections by analyzing ISG expression in RNA-seq data from diverse species, revealing hidden infections missed by conventional methods.

This study introduces a pLM entropy metric to assess protein site conservation and variability, showing that fine-tuning pLMs on evolutionarily related proteins improves their ability to capture context-specific site variation.

This study identifies the non-spike R204P mutation in the SARS-CoV-2 Omicron XEC variant as a driver of increased inflammation and pathogenicity through NF-κB activation.

This study identifies ACE2 diversity in horseshoe bats and shows bidirectional evolutionary shaping between ACE2 and sarbecovirus spike proteins.

This study identifies structural and phenotypic changes in the RBD loop2 region of HKU5r-CoVs as key factors enabling their emergence in mink and potential zoonotic transmission.

Viro3D predicts and curates protein structures for 4,400 viruses, expanding structural coverage 30-fold and enabling functional and evolutionary analysis of viral proteins.

This study investigates the ACE2 tropism of sarbecoviruses across 17 mammals using pseudovirus assays, structural analysis, and phylogenetics to identify genetic factors influencing host range.

This study uses a protein language model to map influenza virus antigenicity from genotype data, revealing evolutionary patterns and improving vaccine strain selection.

This paper introduces CoVFit, a protein language model that predicts SARS-CoV-2 variant fitness using spike protein sequences, identifying fitness-enhancing mutations and predicting emerging variants.

This study expands the known diversity of human anelloviruses by analyzing 829 genomes from microbiome datasets and identifies new clades and recombination patterns.

The paper maps glycoprotein structures across Flaviviridae to identify evolutionary events shaping their virology and ecology.

This paper shows that protein language models can assess mutation effects in SARS-CoV-2 spike protein without multiple sequence alignments, capturing evolutionary constraints and variant properties.

The paper introduces a database-integrated genome screening approach to explore the dark genome, identifying over 1.1 million endogenous viral elements across 874 vertebrate species.

The paper identifies conserved recombination patterns across six coronavirus subgenera and links them to transcriptional regulatory sequences and sequence features.

The paper introduces a new metric, Synonymous Dinucleotide Usage (SDU), to quantify dinucleotide representation in viral genomes and demonstrates its application on flaviviruses and rhabdoviruses.