CoVizu

CoVizu visualizes and analyzes global SARS-CoV-2 genomic diversity to support large-scale genomic epidemiology and variant tracking using genomes from repositories such as GISAID.

Key Features:

• Rapid analysis and quality filtering: Python-based processing filters genomes for problematic sites, incomplete coverage, and excessive divergence from a strict molecular clock.
• Feature extraction and storage: Extracts all differences from the reference genome, including insertions and deletions (indels), using minimap2 and stores them compactly as per-genome features.
• Variant identification and weighting: Within each Pango lineage, collapses genomes with identical feature sets into variants, generates 100 bootstrap samples of the union of feature sets to assign weights, and computes symmetric differences between weighted feature sets for all variant pairs.
• Phylogenetic tree construction: Builds neighbor-joining trees with RapidNJ from the resulting distance matrices and derives a majority-rule consensus tree per lineage, collapsing branches with support below 50% or mean length below 0.5 differences and mapping affected tip labels to internal nodes as directly-sampled ancestral variants.
• Visualization with beadplots: Renders lineage consensus trees as beadplots using D3.js, representing variants as horizontal segments annotated with beads by sample collection date and connecting variants with vertical edges for branches.
• Scalability: Processes approximately one million genomes in about nine hours using 34 cores.

Scientific Applications:

• Genomic surveillance: Track emergence and global spread of SARS-CoV-2 variants using large-scale genome datasets such as GISAID.
• Evolutionary dynamics: Elucidate evolutionary relationships and dynamics among variants within Pango lineages through compact feature-based comparisons and consensus trees.
• Transmission and variant characterization: Support analyses of transmission patterns and variant characteristics by temporally annotating samples on beadplots.

Methodology:

Filter genomes for quality and clock divergence; extract reference differences including indels with minimap2 and store as features; collapse identical-feature genomes into variants per Pango lineage; generate 100 bootstrap samples of the union of feature sets to weight features and compute symmetric differences between weighted feature sets to produce distance matrices; construct neighbor-joining trees with RapidNJ and form majority-rule consensus trees per lineage, collapsing low-support or short branches and mapping affected tips to internal nodes; render consensus trees as beadplots with temporal sample annotation using D3.js.