Spacedust

Spacedust identifies conserved gene clusters across bacterial, archaeal, and viral genomes by combining protein homology and gene-neighborhood conservation to support functional inference in metagenomic and comparative genomics studies.

Key Features:

• Modular Design: Modular architecture for integration into bioinformatics workflows.
• Homology and Neighborhood Conservation: Detects conserved gene clusters by analyzing homologous protein matches alongside conservation of gene neighborhood.
• Structure Comparison and Homology Search: Uses Foldseek for fast structure comparisons and MMseqs2 for sequence homology searches.
• Agglomerative Hierarchical Clustering: Aggregates sets of homologous hits between genome pairs and identifies clusters preserving neighborhood using an agglomerative hierarchical clustering algorithm.
• Order Conservation P-values: Computes order conservation p-values to detect partially conserved clusters and assess conservation significance.

Scientific Applications:

• Metagenomic annotation: Facilitates annotation of sequenced bacterial, archaeal, and viral genomes in metagenomic datasets by identifying conserved clusters.
• Functional association inference: Enables inference of functional associations and biological processes from conserved gene neighborhoods.
• Microbial ecology and evolution: Supports studies of microbial ecology and evolution by revealing conserved genomic modules across genomes.
• Antiviral defense system detection: Detects antiviral defense system clusters, with validation against PADLOC annotations.

Methodology:

Performs de novo discovery of conserved gene clusters via all‑versus‑all genome comparisons; uses Foldseek for structure comparisons and MMseqs2 for homology searches; aggregates homologous hits with an agglomerative hierarchical clustering algorithm and computes order conservation p-values to detect partially conserved clusters; demonstrated in an all‑versus‑all analysis of 1,308 bacterial genomes identifying 72,843 conserved gene clusters containing 58% of the 4.2 million genes analyzed and validated by recovering 95% of PADLOC-annotated antiviral defense system clusters.