GUNC & GENC: Genome Quality Control and Genomic Neighborhood Conservation in Prokaryotes

In prokaryotes, proteins often work together in molecular or cellular functional modules (e.g. complexes or pathways) and their respective genes tend to be co-regulated together within operons. Gene neighborhood information within genomes can therefore be used to infer functional associations between proteins. Such genomic context utilization approaches become both more powerful and challenging with the exponentially growing amount of genomic information from isolate-derived and metagenome-assembled genomes (MAGs), where also quality issues need to be considered. In this thesis, I describe a new tool for quality control of isolate genomes and MAGs and use the improved genomic information as a basis for a computational framework that enables large-scale analysis of genomic neighborhood which I illustrate with selected examples. Metagenomes are sourced from many different microbial habitats and provide a broad and increasingly more comprehensive coverage of microbial lifestyles and environmental adaptations. Metagenomes constitute a mix of genomic sequences of all organisms present in an environmental sample. Therefore, the process of reconstructing complete original genomes by binning assembled metagenomic contigs while avoiding contamination by unrelated contigs remains a decade old technical challenge. Following metagenomic binning, MAGs are filtered for quality based on their genome completeness and contamination provided by the currently most popular quality control tool called CheckM. For that, CheckM uses presence or absence of lineage-specific single-copy marker genes which makes it less sensitive towards nonredundant contamination. To address this issue, and as a prerequisite for genomic context analyses in metagenomes, I developed a tool called GUNC (Genome UNClutterer) described in the first chapter of this thesis. In simple terms, GUNC uses all genes in a genome to score how well contig boundaries separate genes with unrelated lineages which is a sign of contamination. Comprehensive contamination simulations showed that GUNC complements CheckM in detecting contamination in prokaryotic genomes. By GUNC’s conservative estimates 5.7% of genomes in GenBank, 5.2% in RefSeq (both are databases for isolate-derived genomes), and 15-30% of pre-filtered MAGs in recent studies are undetected chimeras containing misbinned contigs. As a result, quality filtering of MAGs using both CheckM and GUNC further improves the quality and reduces contamination. I then use these cleaner xii genomes as input for a novel computational framework for the integration and analysis of GEnomic Neighborhood Conservation (GENC) presented in the second chapter of this thesis. In GENC, the tendency of genes to neighbor each other in operons was integrated and scored at the level of orthologous groups (OGs) of gene products, i.e., proteins, and was used to construct networks of genomic neighborhood conservation (GNC) between OGs for each taxonomic clade across different taxonomic levels based on over 2 million MAGs and isolatederived genomes. These large networks allowed de novo functional annotation of ~20% of all OGs using existing annotations of neighboring genes which were available for ~25% of OGs. Clustering of the GNC networks resulted in functional modules that are a network representation of conserved operons. These functional modules were shown to be frequently enriched in specific phylogenetic clades or microbial habitats. Additionally, the hierarchical relationships between networks corresponding to clades in the tree of life and also between hierarchically built OGs in networks are a powerful tool for investigating variations of GNC between subclades of larger clades. I show how OGs can change their conserved genomic context and potentially also functional context across subclades. 23,251 such cases of functional rewiring were systematically detected across the entire prokaryotic tree of life. As an example, a case of an aminotransferase OG is shown that switches between operons associated with ectoine and siderophore biosynthesis pathways in a polyphyletic and mutually exclusive manner. Such complementation of sequence phylogeny with genomic (and subsequently functional) context conservation patterns provides valuable additional perspective into studying the evolution of function and functional context of proteins. To conclude, GUNC contributes to improving quality control of prokaryotic genomes while GNC networks in GENC and their downstream applications for functional insight can help better utilize quickly expanding microbial genomic neighborhood information for investigating protein function, functional context and evolution across the prokaryotic tree of life.