Genome Detective Coronavirus Typing Tool for rapid identification and characterization of novel coronavirus genomes

Cleemput, Sara; Dumon, Wim; Fonseca, Vagner; Karim, Wasim Abdool; Giovanetti, Marta; Alcantara, Luiz Carlos; Deforche, Koen; de Oliveira, Tulio

cludes 7 sequences from bats which did not cause large human outbreaks. Cluster 3 (named as Bat SARS-CoV HKU3) includes three WGS sampled from Rhinolophus sinicus (i.e. Chinese rufous horseshoe bats). Cluster 4 (Bat SARS-CoV ZXC21/ZC45) includes two SARSr-CoV sampled from Rhinolophus sinicus bats in Zhoushan, China. Cluster 5 (virus named SARS-CoV-2 by the ICTV committee and disease named COVID-19 by the WHO) includes three public sequences from the outbreak. We identified this cluster with many sequences from GISAID but kept only three ones as these were the first GenBank sequences. The first whole genome of SARS-CoV-2 was kindly shared by Prof. Yong-Zhen Zhang and colleagues in the virological.org website. Detailed information about the phylogenetic reference datasets are available in Supplementary Table 2. The phylogenetic reference dataset was used to create an automated Coronavirus Typing Tool using the Genome Detective framework (Vilsker et al., 2019 , Fonseca et al., 2019 . To determine the accuracy of this tool, each of the 431 test WGS was considered for evaluation (i.e. 384 reference sequences from VIPR and 47 public SARS-CoV-2 sequences). The sensitivity, specificity and accuracy of our method was calculated for both species assignment and phylogenetic clustering of Classifying query sequences in an automated fashion involves two steps. The first step enables virus species assignments and the second, which is restricted to SARSr-CoV, includes phylogenetic analysis. The first classification analysis subjects a query sequence to BLAST and AGA analysi. AGA is a novel alignment method for nucleic acid sequences against annotated genomes from NCBI RefSeq Virus Database. AGA (Deforche 2017 ) expands the optimal alignment algorithms of Smith-Waterman (Smith & Waterman 1981) and Gotoh (Gotoh 1982 ) based on an induction state with additional parameters. The result is a more accurate aligner, as it takes into account both nucleotide and protein scores and identifies all of the polymorphisms at nucleotide and amino acid levels. In the second step, a query sequence is aligned against the phylogenetic reference dataset using -add alignment option in the MAFFT software (Katoh & Standley 2013) . In addition, a Neighbor Joining phylogenetic tree is constructed using the HKY distance metric with gamma among-site rate variation with 1,000 bootstrap replicates using PAUP* (Swofford) . The query sequence is assigned to a particular phylogenetic cluster if it clusters monophyletically with that clade or a subset of it with bootstrap support >70%. If the bootstrap support is <70%, the genotype is reported to be unassigned.