Taxonomic profiles from shotgun metagenomes (MGX) and metatranscriptomes (MTX) typically exclude viruses for several reasons: inconsistent viral nucleotide extraction; small viral genome sizes, and subsequently a small proportion of viral genetic content; lack of universal viral marker genes; multiple nucleic acid backbone types; rapid evolution, recombination, and sequence divergence; and most prominently, a lack of well-characterized viral reference databases. Current approaches thus typically require assembly and subsequent classification of viral contigs. Like bacterial metagenome-assembled genomes (MAGs), this approach is extremely successful for viral MAG (vMAG) discovery, but it greatly lacks sensitivity for viral profiling or detection, as only a moderate fraction of MGX/MTX can typically be assembled at all. 
To address this, we developed BAQLaVa (Bioinformatic Application for Quantification and Labeling of Viral taxonomy), which integrates tiered reference-based profiling to generate viral taxonomic profiles from shotgun DNA or RNA sequencing. BAQLaVa profiles viruses across 122,099 Viral Genome Bins (VGBs) generated by integrating both well-characterized viral databases (ICTV and RefSeq) and novel vMAG databases. Our integration process also reconciles putative nomenclature with recognized ICTV taxonomy, as well as resolves highly recombinant viral genomes recalcitrant to genomic similarity clustering. From each resulting VGB, we identify both windows of unique marker nucleotide regions and pan-genome contents to which translated amino acid matches can be made. The resulting databases are then used to generate a harmonized VGB profile from MGX/MTX reads. We evaluated this approach with a set of synthetic genomes representing a spectrum of “known” (within-database) and “novel” viruses, and observe that BAQLaVa is highly accurate, achieving precision and recall >95% for typical strains of known viruses; >71% for highly diverged strains; and a low false positive rate (<0.01%) when exposed to novel viruses.We applied BAQLaVa to 1,626 MGX and 816 paired MTX from the HMP2 Inflammatory Bowel Disease Multi’omics Database (IDBMDB), which longitudinally surveyed 130 individuals with inflammatory bowel disease (IBD) or non-IBD controls. We observed significant changes to the virome in both active and inactive IBD – unlike the bacterial microbiome, which is specifically disrupted during active disease. Both the bacteriome and virome lost diversity during inflammation. Viruses were generally stable within-subject over time, with reduced stability during disease, and 21 of 30 VGBs associated with IBD were novel clusters which did not contain any previously characterized viruses. Strikingly, this is an overrepresentation of novel viruses compared both to the viruses associated with dysbiosis (40% unclassified) and to the BAQLaVa database at large (7% of VGBs unclassified), indicating that these novel viruses may be an important but yet unstudied factor in IBD. BAQLaVa thus provides dramatically improved capabilities for viral detection and quantification from unenriched, unassembled MGX and MTX, which have already provided insight into IBD pathology not previously observed.