{"paper_id":"3d8c5823-cc95-491a-8a15-4e4900ffdbc6","body_text":"N E W R E S U L T S\nAddressing the dynamic nature of reference data: a new\nnt database for robust metagenomic classification\nJose Manuel Martí1,*, Car Reen Kok2, James B. Thissen2, Nisha J. Mulakken1,\nAram Avila-Herrera1, Crystal J. Jaing2, Jonathan E. Allen1 and Nicholas A. Be2,*\n1Global Security Computing Applications Division, Lawrence Livermore National Laboratory , Livermore, California\n94550, USA and 2Biosciences and Biotechnology Division, Lawrence Livermore National Laboratory , Livermore,\nCalifornia 94550, USA\n*Correspondence emails: martimartine1@llnl.gov; be1@llnl.gov\nAbstract\nBackground\nAccurate metagenomic classification relies on comprehensive, up-to-date, and validated reference databases. While the NCBI\nBLAST Nucleotide (nt) database, encompassing a vast collection of sequences from all domains of life, represents an invaluable\nresource, its massive size —currently exceeding 10 12 nucleotides— and exponential growth pose significant challenges for\nresearchers seeking to maintain current nt-based indices for metagenomic classification. Recognizing that no nt-based indices\nexist for the widely used Centrifuge classifier, and the last public version was released in 2018, we addressed this critical gap.\nResults\nWe present a new Centrifuge-compatible nt database, meticulously constructed using a novel pipeline incorporating different\nquality control measures, including reference decontamination and filtering. These measures demonstrably reduce spurious\nclassifications, and through temporal comparisons, we reveal how this approach minimizes inconsistencies in taxonomic\nassignments stemming from asynchronous updates between public sequence and taxonomy databases. These discrepancies are\nparticularly evident in taxa such as Listeria monocytogenes and Naegleria fowleri, where classification accuracy varied significantly\nacross database versions.\nConclusions\nThis new database, made available as a pre-built Centrifuge index, responds to the need for an open, robust, nt-based pipeline for\ntaxonomic classification in metagenomics. Applications such as environmental metagenomics, forensics, and clinical\nmetagenomics, require comprehensive taxonomic coverage and will benefit from this resource. Our new nt-based index\nhighlights the importance of treating reference databases as dynamic entities, subject to ongoing quality control and validation\nakin to software development best practices. This dynamic update approach is crucial for ensuring the accuracy and reliability of\nmetagenomic analysis, especially as databases continue to expand in size and complexity .\nKey words: Metagenomics; Taxonomic Classification; Reference Database; NCBI BLAST nt; Centrifuge; Recentrifuge; Quality\nControl; Data Contamination; High Performance Computing\nBackground\nThe capacity to accurately and sensitively profile microbial commu-\nnities is critical to a broad swath of biological applications, spanning\nbiomedical research, healthcare, and environmental biosurveil-\nlance. Microbial community features are indicative of a multitude\nof human health-relevant outcomes [1], including cancer [2], infec-\ntion [3, 4, 5], trauma [6, 7], chronic injury [8, 9], and neurological\ndisease [10, 11, 12]. Similarly, profiling microbial communities in\nnon-human environments can provide detection and association\nCompiled on: June 12, 2024.\nDraft manuscript prepared by the author.\n1\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\n2 |\nDatabase\nCore processing Post-processing\nDownstream
\npipeline \n(e.g. AI/ML) \nPipeline for Centrifuge
\n database preparation\nPre-processing\ndata  \n \nFASTQ\nmetadata \nRecentrifuge’s interactive \nplot for visualization \nand exploration of results \n  \nHTML\nCentrifuge \nC++\n
\n \npython\nValidation
\nanalysis\nNCBI BLAST nt  \n(FASTA)\nNCBI BLAST nt \n(Centrifuge)\nNCBI \nTaxonomy\nFigure 1. General layout of a metagenomic classification pipeline based on the NCBI BLAST nt database using Centrifuge and Recentrifuge. This flowchart is showing a typical\nconfiguration of a pipeline for taxonomic classification in metagenomics using the NCBI BLAST nt database [37, 39]. Data (and metadata, as available) is feed into a FASTQ\nformat pre-processing block that performs quality checks using software such as fastp [40] and MultiQC [41]. Filtered FASTQ files are then processed by Centrifuge [20] using\nan indexed version of the nt database. Finally, Recentrifuge [36] post-process Centrifuge output, including negative control samples (if any), to further filter and prepare\nresults for interactive exploration, downstream analysis and validation tests. Typical downstream jobs include metagenome assemblers and artificial intelligence/machine\nlearning (AI/ML) training pipelines. The pipeline for building the Centrifuge reference database based on nt is detailed in Figure 3.\ninformation relevant to bioremediation, environmental surveil-\nlance [13, 14], pest control [15], and wastewater monitoring [16, 17].\nIn these applications spaces, the microbial community is often\nprofiled via metagenomic sequencing, which facilitates compre-\nhensive detection of fastidious or difficult-to-culture microorgan-\nisms. Metagenomic classification transforms raw shotgun DNA\nsequences into microbiome composition profiles, which can then\nbe interrogated for signatures or biomarkers of any given label of\ninterest.\nA broad array of software platforms have been developed for\nclassification of shotgun metagenomic sequence reads, including\nKraken [18, 19], Centrifuge [20], MetaPhlAn [21], CLARK [22, 23],\nDIAMOND [ 24], Kaiju [ 25], GOTTCHA [ 26], metakallisto [ 27],\nKMCP [28], LMAT [29, 30], and ganon [ 31]. Each of these plat-\nforms relies, in some form, on a reference database containing\nknown sequences or sequence signatures from a subset of refer-\nence genomes. These databases may include a range of indexed\nreference genomes, a subset of discriminative k-mers (e.g., CLARK\n[22] and KrakenUniq [32]), or a curated set of marker genes (e.g.,\nMetaPhlAn4 [21]). The selection of database and software plat-\nform has a clear and systematic impact on the resultant microbial\nprofile, influencing downstream assessments [33]. Numerous ad-\nditional benchmarking studies providing comparative assessments\nbetween platforms are available for assessment [34]. In addition to\nthe reference source and taxonomic clades for inclusion, metage-\nnomic classification results are impacted by temporal distribution\nof build dates and filtering processes applied to the reference con-\ntent. These parameters are potentially impactful but have not been\nthoroughly assessed in previous studies.\nTo address this knowledge gap, we undertook a comparison\nof four temporally distinct database builds, comparing the extent\nof read classification and overall compositional impact. We also\nassessed the impact of reference decontamination via the Conter-\nminator [35] and Recentrifuge [ 36] platforms and the impact of\nshort reference sequence removal. For broad taxonomic coverage,\nwe constructed databases from the NCBI BLAST Nucleotide (nt)\ndatabase [37], the most comprehensive database for nucleotide\nBLAST search [38]. We selected the Centrifuge classifier platform\nfor this comparison due to its classification speed and optimized\nindexing scheme.\nThe NCBI BLAST nt database encompasses nearly all traditional\nGenBank divisions, representing a significant portion of available\nGenBank sequences and spanning all domains of life [37, 39]. There-\nfore, the growth rate of nt follows that of GenBank. Both the num-\nFigure 2. Evolution of NCBI GenBank database over time. This semi-log plot shows\nthe number of bases and sequences for NCBI GenBank. For the last 3 years of the time\nseries, from April 2021 to April 2024, we fit an exponential model Aϕ(t–t 0)/τ (with\ntime constant τ = 1) to the number of bases in the database, with R2 confirming the\ngoodness of the model. As timet (x-axis, date) is measured in years fromt0 = 1982, A\nis the intercept to t = t0 and ϕ is the base of the exponential with the meaning of the\nyearly rate of change [44]. So, on average, between the mentioned dates, GenBank\ndatabase had a 56% growing rate in nucleotides from year to year. In addition, at\nleast in the last decade, the slope of the number of bases has been increasing while\nthe slope of sequences has been decreasing, with these opposite trends indicating a\nprogressive rise in the average length of the sequences in the database.\nber of sequences and the nucleotides in GenBank are experienc-\ning an exponential growth (see Figure 2 for quantitative details).\nBuilding nt-based reference databases for taxonomic annotation\nrequires increasing computational resources, such as processing\npower, memory, and I/O bandwidth. In particular, the build pro-\ncess for these databases is memory-intensive, requiring comput-\ning architectures that are not broadly available, with the literature\nmentioning initial database build difficulties and subsequent build\nfailures as the dataset sizes increase[31, 42]. The latest available\nCentrifuge database based on NCBI BLAST nt was built in 2018,\nwith multiple users unsuccessfully requesting an updated version\nin recent years [43]. As a response to the need for an open, robust,\nnt-based pipeline for taxonomic classification in metagenomics, we\nprovide an updated, stable, and optimized Centrifuge nt reference\ndatabase.\nThe updated database provides a critical resource for researchers\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\nMartí et al. | 3\nTable 1. Size parameters for nt DB references and builds\nnt DB release date Build version name Masked nt DB size Centrifuge nt DB size Taxonomy DB\nstd dec std dec size content\nJun 26, 2022 Jun22 - 745 GiB - 442 GiB 203 MiB 2.43 Mtid\nSep 23, 2022 Sep22 - 843 GiB - 500 GiB 213 MiB 2.45 Mtid\nJan 04, 2023 Jan23 941 GiB 936 GiB 558 GiB 555 GiB 217 MiB 2.48 Mtid\nApr 03, 2023 Apr23 1.09 TiB 1.08 TiB 615 GiB 611 GiB 220 MiB 2.50 Mtid\nAbout database sizes, dec is a decontaminated database while std is a standard database (non-decontaminated).\nThe size of taxonomy database is the decompressed size in megabytes of the names.dmp file of the corresponding NCBI Taxonomy database available at the indicated nt\ndatabase release date.\nThe unit tid denotes (different) taxonomic identifiers, so Mtid is a million of taxonomic identifiers.\nusing metagenomics, where comprehensive and reliable taxonomic\nclassification across the tree of life is paramount. This work un-\nderscores the critical need for treating the databases as a dynamic\nentity requiring continuous quality control and validation, much\nlike software development best practices.\nMethods\nReference selection\nFour NCBI nt release dates spanning 2022-2023 were selected for\nassessment (Table 1).\nCharacterization and decontamination of the reference\ndatabase\nFigure 3 shows a flowchart of the main steps involved in the pipeline\nfor building a reference database for Centrifuge based on the nt\ndatabase.\nAfter obtaining the nt database in FASTA format using BLAST+\ntools [45] and the current content of the NCBI Taxonomy database,\nthe content of nt was explored using Recentrifuge [36]. Figure S1\nshows some examples of insights into the database provided at this\nstep.\nThe nt database was decontaminated using Conterminator [35]\nin parallel and the output post-processed by a different script of\nRecentrifuge, which removes the detected contamination and gen-\nerated a contamination network plot (see Figure S2 for an example).\nReference masking, filtering, and indexing\nTherefasplit script from Recentrifuge [36] was used at different\nstages of the pipeline to enable efficient I/O processing of the se-\nquence database in parallel, which contains nucleotide counts on\nthe order of 1012. Low complexity regions were masked using sev-\neral instances of NCBI dustmasker [45] in parallel. The decontami-\nnated and masked version of the nt database was then processed by\nRecentrifuge’srefafilt code to remove sequences shorter than 16\nnucleotides (found to cause problems downstream) and demultiplex\nheaders for redundant sequences as generated by recent versions\nof the BLAST+ command line toolblastdbcmd. Finally , indexing of\nthe decontaminated filtered database was performed with an opti-\nmized version of the Centrifuge 1.0.4-β database building C++ code\ncentrifuge-build [20] targeting Livermore Computing (LC) HPC\ncluster nodes with 128 cores and 2 TiB main memory. The entire\nbuild process required about three weeks of clock time.\nSample sequencing and processing\nDefined DNA mixtures were employed for comparative assess-\nment of databases and classification parameters (Tables 1, 2). No-\ntemplate process negative controls (ntc) were included in all analy-\nses. Sequence libraries were prepared from gDNA via the Illumina\nDNA Prep library preparation kit and sequenced on the Illumina\nNextSeq 2000 via P2 300 cycle reagents and flow cell. Resultant\ndata were processed for taxonomic classification via Centrifuge us-\ning each of the database builds described in Table 1. Comparative\nanalyses of reference controls were carried out in R (version 4.2.0).\nPhyloseq [46] was used to generate microbiome abundance profiles\nfrom count data and ggplot2 [47] was used to generate figures. The\nvegan package [48] was used to compute Bray Curtis dissimilarity\n \nreconterminator \npython\ncentrifuge-build C++\nNCBI nt \n(FASTA)\nNCBI nt \n (Centrifuge)\nNCBI \nBLAST nt\nBLAST+ tools \nC++\ndustmasker \nC\ndustmasker \nC\ndustmasker \nC\ndustmasker \nC++\nNCBI \ntaxonomy\nNCBI nt \n(decontaminated)\nNCBI nt \n(conterminated)NCBI nt \n(conterminated)NCBI nt \n(conterminated)NCBI nt \n(decontaminated)\n \ntaxonomic \ndata\nconterminator \nC++\n \nConterminator \noutput \ntsv\ncontamination \nnetwork \nPDF\nNCBI nt \n(dustmasked)\n \nrefasplit \npython\n \nrefafilt \npython\nlength/reads \nlog-histogram \nPDF\nNCBI nt \n(filtered)\n \nrcf \npython\nseqs/len per taxon \n(krona-style) \nHTML\nLEGEND \nfile\nfile \nformat\nRequired input \n(multiple files)\nResults of different analysis \nthrough the pipeline \n(one single file per run) Data flow\ncode\nlang/env Pieces of software\ninput \nformat\nNCBI BLAST \nnt database\noutput \nformat\nfile Intermediate results \n(multiple files)\nFigure 3. General flowchart of the pipeline for building the Centrifuge reference database based on the NCBI BLAST nt database. The entire pipeline process takes several\nweeks using a HPC (high-performance computing) server with a high-memory-per-core ratio, a high-bandwidth low-latency parallel filesystem, 128 cores, and total\nmemory of 2 TiB. The most demanding step is the final indexing of the filtered database using a version ofcentrifuge-build, which may expand beyond two weeks, followed\nby the decontamination step, which usually takes a few days.\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\n4 |\nTable 2. DNA standard mixtures employed for comparative assessment\nSample name Mixture Vendor Cat No Human spike Sample type\nzymo_gut_microbiome Zymobiomics gut microbiome Zymo D6331 no bacteria\nzymo_micro_comm_II Zymobiomics microbial community standard II Zymo D6311 no bacteria\natcc_mycobiome Mycobiome mock community ATCC MSA-1010 no eukaryote\natcc_oral_spike Oral microbiome mock community ATCC MSA-1004 18 ng bacteria\natcc_skin_spike Skin microbiome mock community ATCC MSA-1005 14 ng bacteria\natcc_gut_spike Gut microbiome mock community ATCC MSA-1006 40 ng bacteria\nprotozoa_mix Acanthamoeba castellanii, Naegleria fowleri ATCC Various no eukaryote\nmammal_equal_mix Equal mix of 11 mammalian DNAs Various Various no eukaryote\nnon-mammal_mix Equal mix of 10 non-mammalian eukaryotic DNAs Various Various no eukaryote\nmammal_log_mix Log concentration range of 7 mammalian DNAs Various Various no eukaryote\natcc_20_strain ATCC 20 strain mix ATCC MSA-1003 no bacteria\natcc_path_mix ATCC pathogen mix ATCC MSA-4000 no bacteria\ndistances between samples.\nResults and Discussion\nComparative assessment of gDNA reference controls and\nmixtures\nIn addition to the assessment of temporal database builds, we eval-\nuated other microbiome analytical parameters relevant to the as-\nsessment of accurate microbiome profiles. This warrants the use\nof standard controls with known microbial composition across dif-\nferent kingdoms of life (Tables 1, 2). We systematically evaluated\nthe impact of different values for the minimum length of partial\nhits in Centrifuge [\n20] or MHL (Minimum Hit Length), relative\nabundance thresholds, taxonomic levels, and sample type to gener-\nate accurate microbiome profiles. We calculated precision, recall,\nbalanced F1 scores and Bray Curtis dissimilarity distances to com-\npare sequenced profiles of control standards with ground truth data\nacross all database versions (Table 2).\nFirst, we investigated the impact of temporal database builds\non microbiome profiling and observed temporal instability in Bray\nCurtis dissimilarity distances in 3 samples; ATCC_path_mix, proto-\nzoa_mix and zymo_micro_comm_II when compared to reference\ncontrol profiles (Figures 4, S3). There was an increase in Bray Cur-\ntis dissimilarity distances in the Jan23 database versions for these\nsamples. Upon inspection, we determined that delayed updates be-\ntween taxonomic labels in NCBI BLAST nt database and the current\nNCBI Taxonomy database at the time of the nt database release was\na major factor. The nt database is relying on the Taxonomy database\nfor the taxonomic information, but both databases are evolving over\ntime independently, with no explicitly version mapping between\nthem in the documentation. Traditionally , the nt database was us-\ning a current version of the taxonomy database but that does not\nseem the case anymore, with the taxonomy used in NCBI BLAST nt\ndatabase slightly trailing behind. In some cases, newly uploaded\ngenomic sequences were lacking current taxonomic labels and then\nconfounded Centrifuge’s database indexing algorithm and, eventu-\nally , its classification process.\nFor example, in the protozoa_mix sample, there were reads\nthat were assigned and equally scored to multiple Naegleria fowleri\n(tid 5763) reference genomes that were present in the Jan23 NCBI\nBLAST nt database. However, when Centrifuge was reporting multi-\nple assignments (Centrifuge default), only reads that were assigned\nto N. fowleri genomes whose tid was present in the NCBI Taxon-\nomy database used during the database compilation were assigned\nto tid 5763 while reads that were assigned to N. fowleri genomes\nwhose tid was absent from the used taxonomy (but present in the\nNCBI BLAST nt database) were classified with a tid 0, normally re-\nserved for unclassified reads. When Centrifuge was configured to\nreport a unique tid per read using the lowest-common-ancestor\n(LCA) strategy (pipeline default), those reads were given a final\nroot-only assignment (tid 1). As another example, the results for\nthe zymo_micro_comm_II sample were impacted when using the\nApr23 database due to varied taxonomic classification of Listeria\nmonocytogenes genomes (Figures 4B, S3). In that databse, the addi-\ntion of a Listeria monocytogenes genome (CP046449.1) annotated as\nListeria sp. LM90SB2 (tid 2678528 in the NCBI Taxonomy database,\nwith parent being unclassified Listeria, tid 2642072) resulted in a\nfinal LCA classification of reads at only the genus level ( Listeria).\nOverall, our observations highlight the impact on the classification\nof specific taxa due to the discrepancy of taxonomic information\nbetween the nt database and the version of the NCBI Taxonomy\ndatabase used for indexing the Centrifuge database.\nNext, we tested the impact of the following MHL values: 15\n(minimum value), 22 (Centrifuge’s default), 30, 40, and 50 on the\nnumber of unique NCBI tax ids assigned to sequenced reads. Here,\nwe observed that the number of unique NCBI tax id assignments\nand Bray Curtis dissimilarity distances decreases at increasing MHL\nvalues (Figure 4A,B). An inflection point was observed at MHL 30\n(Figure 4A), followed by an increase in F1 and precision scores com-\npared to MHL 15 and 22 (Figure 4D). This suggests that a MHL of\n30 should be close to the optimal value in reducing false positive as-\nsignments across these particular samples (Figure S3B) by reducing\nthe noise of spurious assignments without a significant increase in\nfalse negatives.\nAs most microbiome analysis protocols remove low-abundance\nfeatures to account for potential sequencing errors and false pos-\nitives, we assessed the impacts of different thresholds (0, 0.001,\n0.01, 0.1, 0.5, 1) on classification performance. Overall, we observed\nthat a relative abundance threshold of 0.1% resulted in the highest\nF1 scores across samples (Figure 4D). As expected and previously\nobserved, our data demonstrated that classification performance\ndecreases at lower taxonomic ranks with an increasing number of\ndetected false positives (Fig. S3B). Furthermore, samples consisting\nmainly of bacterial content had a better classification performance\nthan eukaryotic samples (Figure 4C,D).\nFinally, we compared the number of annotated sequences be-\ntween the different versions of the Jan23 database and observed\na decrease along the implemented improvements (Fig 4C). The\nstandard version (Jan23_std) had an average number of 26,309,720\nannotated sequences, followed by the decontaminated version\n(Jan23_dec) with 26,298,931 annotated sequences and lastly , the de-\ncontaminated and filtered version (Jan23_filt) had 26,219,236 anno-\ntated sequences. However, Jan23_dec and Jan23_filt databases had\nhigher non-root to root-only assignment ratios, suggesting that\nadditional annotated reads with the standard version (Jan23_std)\nwere assigned to the root-only level.\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\nMartí et al. | 5\n0 K\n10 K\n20 K\n30 K\n20 30 40 50\nMHL\nno. of unique NCBI tax ids\nsample\nATCC_20_strain\natcc_gut_spike\natcc_mycobiome\natcc_oral_spike\nATCC_path_mix\natcc_skin_spike\nmammal_equal_mix\nmammal_log_mix\nnon−mammal_mix\nprotozoa_mix\nzymo_gut_microbiome\nzymo_micro_comm_II\n0.1\n0.2\n0.3\n0.4\n0.5\nmhl15 mhl22 mhl30 mhl40 mhl50\nMHL\nBC distance\n0.1\n0.2\n0.3\n0.4\nJun22_dec Sep22_dec Jan23_dec Jan23_std Jan23_filt Apr23_decon Apr23_std\ndatabase\nBC distance\n0.0\n0.1\n0.2\n0.3\n0.4\n0.5\nphylum class order family genus species\ntaxonomic rank\nBC distance\n0.1\n0.2\n0.3\n0.4\nbacteria eukaryote\nsample type\nBC distance\n26,309,720 26,298,931 26,291,236\n1e+05\n1e+06\n1e+07\n1e+08\nJan23_std Jan23_dec Jan23_filt\ndatabase\nno. of classified sequences\n118.98 124.768 124.421\n2\n3\n5\nJan23_std Jan23_dec Jan23_filt\ndatabase\nlog10 (non−root/root only assignments)\nF1\nprecision\nrecall\n0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00\n0.00\n0.25\n0.50\n0.75\n1.00\nmhl\nmhl15\nmhl22\nmhl30\nmhl40\nmhl50\n0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00\n0.25\n0.50\n0.75\n1.00\ndatabase\nJun22_dec\nSep22_dec\nJan23_dec\nJan23_std\nJan23_filt\nApr23_dec\nApr23_std\n0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00\n0.00\n0.25\n0.50\n0.75\n1.00\ntax_rank\nphylum\nclass\norder\nfamily\ngenus\nspecies\n0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00\n0.25\n0.50\n0.75\n1.00\ntype\nbacteria\neukaryote\nA\nB\nC\nD\nFigure 4. Comparative assessment of gDNA reference controls. A) The number of unique NCBItid captured at different MHL values across samples and averaged across all\ntemporal database versions. Data points are colored according to sample. B) Comparison of Bray Curtis dissimilarity distances at an abundance threshold of 0.1% between\nMHL values, database versions, taxonomic ranks, and sample types. For every variable that is being compared, Bray Curits dissimilarity distances were averaged across other\nvariables. C) The number of annotated sequences and ratio of non-root to root-only assigned reads across different versions of the Jan23 database. Mean values are annotated\nat the top of each boxplot. D) Comparison of classification accuracy metrics (F1 balanced accuracy , precision, and recall) between MHL values, database versions, taxonomic\nranks, and sample types across abundance thresholds.\nPotential implications\nWhile this work primarily focuses on providing a readily usable\nand improved nt-based database for metagenomic classification\nwith Centrifuge and Recentrifuge, the underlying methodology\nand findings have implications that extend beyond this immediate\napplication.\nFirstly , the availability of a comprehensive, decontaminated nt\ndatabase opens new avenues for developing and training machine\nlearning models for metagenomic analysis. Such models, trained\non cleaner data, have the potential to achieve higher accuracy and\nbetter generalization capabilities, particularly for challenging tasks\nsuch as identifying novel or rare taxa. This could lead to improved\ntools for metagenome-based diagnostics, discovery of novel micro-\nbial enzymes, and understanding the complex interactions within\nmicrobial communities.\nSecondly , our observation of temporal inconsistencies in taxo-\nnomic assignments stemming from asynchronous updates between\nNCBI databases highlights a crucial challenge in metagenomic anal-\nysis. It underscores the need for a paradigm shift in how we ap-\nproach reference data, moving from static resources to dynamically\nmaintained and versioned collections. Drawing inspiration from\nsoftware development best practices, continuous integration and\ncontinuous delivery (CI/CD) pipelines could be adapted for refer-\nence databases, ensuring timely updates, robust validation, and\nimproved reproducibility across studies.\nAdditionally , the pipeline developed for database construction,\nencompassing decontamination, filtering, and validation steps, can\nbe generalized to other large reference databases commonly em-\nployed in metagenomic analysis. Applying similar quality control\nmeasures to resources like the NCBI nr (non-redundant protein)\ndatabase or specialized databases for viral or fungal identification\ncould significantly enhance the accuracy and reliability of classifi-\ncation in those domains. This has the potential to improve research\noutcomes in fields as diverse as environmental monitoring, disease\ndiagnostics, and evolutionary biology .\nFurthermore, the availability of a comprehensive, up-to-date,\nand validated database encompassing all domains of life—such as\nthe one presented here—offers a unique advantage for researchers\nin fields that require analysis of complex microbial communities\nacross diverse environments. For example, studies investigating\nhost-microbe interactions, exploring the microbiome of underex-\nplored ecosystems, or tracing the origins of emerging pathogens\ncan benefit greatly from such a resource.\nHowever, the exponential growth of sequencing data and the\nassociated computational burden highlight the need for innovative\nstrategies to ensure the long-term sustainability of such compre-\nhensive resources. This might involve exploring distributed com-\nputing approaches, developing more efficient indexing and search\nalgorithms, splitting the database, or even dividing the classifica-\ntion problem, e.g., converting the classification goal in a hierarchi-\ncal problem by using a \"forest\" of specific classifiers to decide the\nannotation at different taxonomic levels instead of a single classifier\ncovering the entire tree of life.\nData availability\nAs of May 2024, the Jan23 decontaminated and filtered Centrifuge\nnt database is the updated database version. It can be downloaded\nfrom LLNL Green Data Oasis (GDO) public ftp server at ftp://\ngdo-bioinformatics.ucllnl.org/centrifuge/nt_wntr23 using any\nsoftware supporting anonymous ftp download. To ease the down-\nload process, the database is split in 71 ultra-compressed7z files of\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\n6 |\n4 GiB or less with name formatnt_wntr23_filt.cf.7z.*. The cor-\nresponding taxonomy files for that version of the database, useful\nfor Recentrifuge runs, are available atftp://gdo-bioinformatics.\nucllnl.org/centrifuge/nt_wntr23/taxonomy.\nShotgun metagenomic sequences of standard controls are up-\nloaded to NCBI SRA with the Bioproject ID PRJNA1118346.\nPractical notes\nDespite the most demanding step is the indexing of the nt database\nas described above, even running the Centrifuge classifier with\nsuch a large indexed database has some important computational\nrequirements, especially regarding free disk space and available\nmain memory. The size of the ultra-compressed database files\namount to 284 GiB. Once decompressed from this format, the size\nof the Centrifuge (compressed) database files is 422GiB. Given that\nsize, a computer with at least half TiB of main memory is recom-\nmended for running the Centrifuge classifier, especially when using\na high number of parallel threads. In addition, Centrifuge may take\nmore than half an hour just for loading the nt database from disk\ninto memory . Given this overhead, it is recommended to use a sam-\nple sheet file (–sample-sheetargument in Centrifuge) to process\nmultiple samples without requiring additional database loads. Also,\nwhen replicating our pipeline, it is strongly recommended to use\nthe LCA strategy with Centrifuge, by using the-k 1 argument in\nthe call to the classifier. Finally, it is straightforward to use Re-\ncentrifuge to post-process Centrifuge’s results, with detailed doc-\numentation in Recentrifuge’s wiki athttps://github.com/khyox/\nrecentrifuge/wiki/Running-recentrifuge-for-Centrifuge.\nDeclarations\nList of abbreviations\nDB: Data Base; HPC: High Performance Computing; I/O: In-\nput/Output; LCA: Lowest Common Ancestor; MHL: Minimum Hit\nLength; WGS: Whole-Genome Shotgun;\nCompeting Interests\nThe authors declare that they have no competing interests.\nFunding\nThis study was supported by the Lawrence Livermore National Lab-\noratory’s Laboratory Directed Research and Development Program\n(LDRD). This work was performed under the auspices of the U.S.\nDepartment of Energy by Lawrence Livermore National Laboratory\nunder Contract DE-AC52-07NA27344. LLNL Disclaimer: Neither\nthe United States government nor Lawrence Livermore National\nSecurity , LLC, nor any of their employees make any warranty, ex-\npressed or implied, or assume any legal liability or responsibility\nfor the accuracy, completeness, or usefulness of any information,\napparatus, product, or process disclosed, or represent that its use\nwould not infringe privately owned rights. Reference herein to any\nspecific commercial product, process, or service by trade name,\ntrademark, manufacturer, or otherwise does not necessarily consti-\ntute or imply its endorsement, recommendation, or favoring by the\nUnited States government or Lawrence Livermore National Security ,\nLLC. The views and opinions of authors expressed herein do not\nnecessarily state or reflect those of the United States government or\nLawrence Livermore National Security , LLC, and shall not be used\nfor advertising or product endorsement purposes.\nReferences\n1. Martí JM, Martínez-Martínez D, Rubio T, Gracia C, Peña\nM, Latorre A, et al. Health and Disease Imprinted in the\nTime Variability of the Human Microbiome. mSystems\n2017;2(2):10.1128/msystems.00144–16.\nhttps://doi.org/10.\n1128/msystems.00144-16.\n2. Sepich-Poore GD, et al. The microbiome and human cancer.\nScience 2021;371(6536).\n3. Johnson RC, et al. Correlation between nasal microbiome com-\nposition and remote purulent skin and soft tissue infections.\nInfection and Immunity 2015;83(2):802–811.\n4. Karl JP . Gut Microbiota-targeted Interventions for Reducing\nthe Incidence, Duration, and Severity of Respiratory Tract In-\nfections in Healthy Non-elderly Adults. Military Medicine\n2021;186(3-4):e310–e318.\n5. van der Lelie D, Taghavi S. COVID-19 and the Gut Microbiome:\nMore than a Gut Feeling. mSystems 2020;5(4).\n6. Avila-Herrera A, et al. Metagenomic features of bioburden\nserve as outcome indicators in combat extremity wounds. Sci-\nentific Reports 2022;12(1):13816.\n7. Kok CR, et al. Targeted metagenomic assessment reflects criti-\ncal colonization in battlefield injuries. Microbiology Spectrum\n2023;11(6):e0252023.\n8. Kalan L, et al. Redefining the Chronic-Wound Microbiome:\nFungal Communities Are Prevalent, Dynamic, and Associated\nwith Delayed Healing. mBio 2016;7(5).\n9. Misic AM, Gardner SE, Grice EA. The Wound Microbiome:\nModern Approaches to Examining the Role of Microorganisms\nin Impaired Chronic Wound Healing. Advances in Wound Care\n2014;3(7):502–510.\n10. Alhasson F, et al. Altered gut microbiome in a mouse model\nof Gulf War Illness causes neuroinflammation and intesti-\nnal injury via leaky gut and TLR4 activation. PLoS ONE\n2017;12(3):e0172914.\n11. McGrattan AM, et al. Diet and Inflammation in Cognitive\nAgeing and Alzheimer’s Disease. Current Nutrition Reports\n2019;8(2):53–65.\n12. Tremlett H, et al. The gut microbiome in human neurological\ndisease: A review. Annals of Neurology 2017;81(3):369–382.\n13. Brodie EL, et al. Urban aerosols harbor diverse and dynamic\nbacterial populations. Proceedings of the National Academy of\nSciences of the United States of America 2007;104(1):299–304.\n14. Jaing C, et al. Sierra Nevada sweep: metagenomic measure-\nments of bioaerosols vertically distributed across the tropo-\nsphere. Scientific Reports 2020;10(1):12399.\n15. Martí JM, Arias-Giraldo LF, Díaz-Villanueva W, Arnau V,\nRodríguez-Franco A, Garay CP . Metatranscriptomic dynamics\nafter Verticillium dahliae infection and root damage in Olea\neuropaea. BMC plant biology 2020;20:1–15. https://doi.org/\n10.1186/s12870-019-2185-0.\n16. Garner E, et al. Next generation sequencing approaches\nto evaluate water and wastewater quality. Water Research\n2021;194:116907.\n17. Keshaviah A, et al. Wastewater monitoring can anchor\nglobal disease surveillance systems. Lancet Global Health\n2023;11(6):e976–e981.\n18. Wood DE, Salzberg SL. Kraken: ultrafast metagenomic se-\nquence classification using exact alignments. Genome biology\n2014;15:1–12.\n19. Wood DE, Lu J, Langmead B. Improved metagenomic analysis\nwith Kraken 2. Genome biology 2019;20:1–13.\n20. Kim D, et al. Centrifuge: rapid and sensitive classification of\nmetagenomic sequences. Genome Research 2016;26(12):1721–\n1729.\n21. Blanco-Miguez A, et al. Extending and improving metage-\nnomic taxonomic profiling with uncharacterized species using\nMetaPhlAn 4. Nature Biotechnology 2023;41(11):1633–1644.\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\nMartí et al. | 7\n22. Ounit R, Wanamaker S, Close TJ, Lonardi S. CLARK: fast and\naccurate classification of metagenomic and genomic sequences\nusing discriminative k-mers. BMC genomics 2015;16:1–13.\n23. Ounit R, Lonardi S. Higher classification sensitivity of\nshort metagenomic reads with CLARK-S. Bioinformatics\n2016;32(24):3823–3825.\n24. Buchfink B, Xie C, Huson DH. Fast and sensitive protein align-\nment using DIAMOND. Nature methods 2015;12(1):59–60.\n25. Menzel P , Ng KL, Krogh A. Fast and sensitive taxonomic classi-\nfication for metagenomics with Kaiju. Nature Communications\n2016;7:11257.\n26. Freitas TA, et al. Accurate read-based metagenome characteri-\nzation using a hierarchical suite of unique signatures. Nucleic\nAcids Research 2015;43(10):e69.\n27. Schaeffer L, et al. Pseudoalignment for metagenomic read\nassignment. Bioinformatics 2017;33(14):2082–2088.\n28. Shen W, et al. KMCP: accurate metagenomic profiling of both\nprokaryotic and viral populations by pseudo-mapping. Bioin-\nformatics 2023;39(1).\n29. Ames SK, et al. Scalable metagenomic taxonomy classifi-\ncation using a reference genome database. Bioinformatics\n2013;29(18):2253–2260.\n30. Ames SK, Gardner SN, Marti JM, Slezak TR, Gokhale MB,\nAllen JE. Using populations of human and microbial genomes\nfor organism detection in metagenomes. Genome research\n2015;25(7):1056–1067.\nhttps://doi.org/10.1101/gr.184879.\n114.\n31. Piro VC, Dadi TH, Seiler E, Reinert K, Renard BY. ganon:\nprecise metagenomics classification against large and up-\nto-date sets of reference sequences. Bioinformatics 2020\n07;36(Supplement_1):i12–i20.\nhttps://doi.org/10.1093/\nbioinformatics/btaa458.\n32. Breitwieser FP , Baker DN, Salzberg SL. KrakenUniq: confi-\ndent and fast metagenomics classification using unique k-mer\ncounts. Genome Biology 2018;19(1):198.\n33. Xu R, Rajeev S, Salvador LCM. The selection of software and\ndatabase for metagenomics sequence analysis impacts the out-\ncome of microbial profiling and pathogen detection. PLoS ONE\n2023;18(4):e0284031.\n34. McIntyre ABR, Ounit R, Afshinnekoo E, et al. Comprehen-\nsive benchmarking and ensemble approaches for metagenomic\nclassifiers. Genome Biology 2017;18(1):182.\n35. Steinegger M, Salzberg SL. Terminating contamination: large-\nscale search identifies more than 2,000,000 contaminated en-\ntries in GenBank. Genome Biology 2020;21(1):115.\n36. Marti JM. Recentrifuge: Robust comparative analysis and\ncontamination removal for metagenomics. PLoS Compu-\ntational Biology 2019;15(4):e1006967.\nhttps://doi.org/10.\n1371/journal.pcbi.1006967.\n37. Coordinators NR. Database resources of the National Center\nfor Biotechnology Information. Nucleic Acids Research 2012\n11;41(D1):D8–D20. https://doi.org/10.1093/nar/gks1189.\n38. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Ba-\nsic local alignment search tool. Journal of molecular biology\n1990;215(3):403–410.\n39. Ye J, McGinnis S, Madden TL. BLAST: improvements\nfor better sequence analysis. Nucleic acids research\n2006;34(suppl_2):W6–W9.\n40. Chen S. Ultrafast one-pass FASTQ data preprocessing,\nquality control, and deduplication using fastp. iMeta\n2023;2(2):e107.\nhttps://onlinelibrary.wiley.com/doi/abs/\n10.1002/imt2.107.\n41. Ewels P , Magnusson M, Lundin S, Käller M. MultiQC: sum-\nmarize analysis results for multiple tools and samples in a sin-\ngle report. Bioinformatics 2016 06;32(19):3047–3048. https:\n//doi.org/10.1093/bioinformatics/btw354.\n42. Plyusnin I, Vapalahti O, Sironen T, Kant R, Smura T. Enhanced\nViral Metagenomics with Lazypipe 2. Viruses 2023;15(2).https:\n//www.mdpi.com/1999-4915/15/2/431.\n43. Chorlton SD. Ten common issues with reference sequence\ndatabases and how to mitigate them. Frontiers in Bioinformat-\nics 2024;4:1278228.\n44. Martí JM, Garay CP . Not just BLAST nt: WGS database joins the\nparty . BioRxiv 2019;p. 653592.\n45. Camacho C, et al. BLAST+: architecture and applications. BMC\nBioinformatics 2009;10:421.\n46. McMurdie PJ, Holmes SP . An R Package for Reproducible Inter-\nactive Analysis and Graphics of Microbiome Census Data. PLoS\nONE 2013;8(5):e61217.\n47. Wickham H. ggplot2: Elegant Graphics for Data Analysis.\nSpringer-Verlag New York; 2016.\n48. Oksanen J, Simpson GL, Blanchet FG, Kindt R, Legendre P ,\nMinchin PR, et al. vegan: Community Ecology Package; 2024,\nr package version 2.6-7.\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\n8 |\nBacteriaBacteria\nPseudomonadota\nGamma...teria\nEnterobacterales\n   1%\nP...s\nPseudomonadaceae\nPseudomonas\n   1%\nXanthomonadales\nXanthomonadaceae\n   1%\nM...s\nMoraxellaceae\n   0.8%\n7 more\nArenicellales\n   0.8%\nAlphaproteobacteria\nHyphomicrobiales\nRhizo...aceaeRhizobi...m groupAgrobacteriumAgrobacteri...ens complex\nAgrobacterium radiobacter\n   7%\n4 more\nSt...aeR...men...es\nuncultured Roseibium sp.\n   1%\n3 more\nRickettsiales\nA...eW...eWo...iaunclassi...olbachia\nWolbachia endosymbiont of Cotesia melitaearum\n   0.9%8 more\nRhod...ales\nParacoccaceae   2%\nRoseobacteraceae   2%\nSphin...dales\nSphingomonadaceae   1%1 more\nRhodospirillales   0.9% uncla...teria\nAlphaproteobacteria bacterium CMAR1   0.9%\nBetaproteobacteria\nBurkholderiales   1%\nTerrabacteria groupBacillota\nBacilliBac...lesB...e\nBacillus   0.9%\n7 more\nP...e\nPaenibacillus\n   0.7% ClostridiaE...s\n2 more\nE...sC...a\nunclassi\n\"ed Candidatus Limiplasma\nCandidatus Limiplasma sp.\n   0.8%\nL...sV...eu...e\nVallitaleaceae bacterium\n   1%\n1 more\nCh...esChr...eaeuncla...aceae\nChristensenellaceae bacterium\n   3%\nNegat...cutes\nSelenomonadales\n   0.7%\nActi...tota\nActi...etes...St...ae\nStreptomyces\n   0.9%\n...\nMicrobacteriaceae\n   0.9%\n5 more7 more\nAci...iia\nA...s\nu...s\nAcidimicrobiales bacterium\n   1%\nCyan...roup\nCy...ta\nC...e\n...NostocaceaeDesmonostoc\nDesmonostoc sp.\n   0.6%\n3 more\nMyco...tota\nMo...es\n...\n...\nPhytoplasma\n   0.7%\nChlo...xota\nAn...ae\n...\n...\n0.9%\n   Aggregatilineales bacterium\nCa...ae\nCal...les\nCal...eae\nunclassif...lineaceae\n4%   Caldilineaceae bacterium\nFCB group\nBacteroid...ota group\nBacteroidota\nFla...iia\nFl...es\n1%   Flavobacteriaceae\nBact...idia\nBa...es\n1 more\n......\n1%   Bacteroidaceae bacterium\n......\n1%   uncultured Microbacter sp.\nMar...les......\n1%   uncultured Saccharicrinis sp.\n......\n0.8%   Marini\n\"laceae bacterium\n4 more\nIgnavi...eriota\nIgnavibacteriaIgn...lesMeli...ceaeunclas...raceae1%   Melioribacteraceae bacterium\nCandidatus ...cibacterotaunclassi\n\"...ibacterota\n1%   Candidatus Latescibacterota bacterium3 more\nPVC groupPla...ota\nPlanct...ycetiaPir...les\nPi...aeMa...us\nunclassi...iblastus\n1%   Mariniblastus sp.\nT...eun...ae3%   Thermoguttaceae bacterium\nPh...aeS...sun...es2%   Sedimentisphaerales bacterium\n... 1%   Phycisphaerae bacterium\nVer...otaOpi...tae...... 0.8%   Opitutales bacteriumLen...otaLe...ia...... 0.7%   Victivallales bacterium\nSpirochaetotaSpi...tiaSpi...lesT...eu...e1%   Treponemataceae bacterium\n1 more\nS...eu...e2%   Sphaerochaetaceae bacteriumuncl...etia\n1%   Spirochaetia bacterium\nThermo...eriotaDesu...adia\nDesu...alesG...eu...e\n2%   Geopsychrobacteraceae bacterium\nDesulf...rioniaDesu...alesD...eP...o\nenvironmental samples0.8%   uncultured Pseudodesulfovibrio sp.\n2 more\n5 more\nAci...ota\nTerr...obia\nuncla...lobia\n1%   Terriglobia bacterium\nVic...ria\nVic...les\nun...es\n4%   Vicinamibacterales bacterium\nBacteria incertae sedisB...a\nC...a\n...\n0.7%\n   Candidatus Poribacteria bacterium\nCan...ota...\n...\n1%   Candidatus Binatia bacterium\nSynergistota\nSyn...tia\nS...s\n...\n...\n0.9%\n   Aminobacteriaceae bacterium\nMyx...ota\nu...a\n1%   Myxococcota bacterium\nLength (log10)\n1.0\n2.4\n3.8\n5.1\n6.5\n7.9\nrootroot\nEukaryota\nOpisthokonta\nMetazoaEumetazoa\nCaniformia\n   2%\nFelidae\n   0.9%\nOdontoceti\n   0.7%\nBovidae\n   0.7%\nMicrochiroptera\n   0.9%\n3 moreHomo sapiens\n   1%\nCercopithecinae\n   0.7%Mus\n   0.6%\nCricetidae\n   0.8%\n5 more\nPasseriformes\n   2%Galliformes\n   0.6%28 more\nDurocryptodira\n   0.7%Toxicofera   0.8%\nAnura   0.8%\nPerciformes   1%\n17 more\nCyprinodontoidei   0.6%\n4 more\nSyngnathinae   0.6%\nCarangaria   0.7%\n5 more\nOncorhynchus\n   0.6%\nCyprininae\n   0.7%\nCharaciphysae\n   0.8%\nmelanogaster group\n   0.6%\nCulicidae\n   0.8%\nAnthophila\n   0.7%\nFormicidae\n   0.6%\n7 more\nPapilionoidea\n   1%\n15 morePolyphaga\n   0.8%\nTimema californicum\n   0.6%\n14 moreHemiptera\n   0.7%\nEumalacostraca\n   0.7%\n0.9%\n   Arachnida\n0.9%\n   Autobranchia\n0.7%\n   Gastropoda\n1%   Platyhelminthes\n3 more\nFungi\nDikarya\n0.9%\n   Hypocreales6 more\n0.7%\n   Aspergillus20 more\n0.6%\n   Pleosporomycetidae\n0.9%\n   Agaricomycetes\n0.7%\n   Fungi incertae sedis\nViridiplantaeStreptophytaStreptophytina\n1%   NPAAA clade\n0.8%   Rosales\n0.6%   Malpighiales5 more\n0.9%   Brassicaceae\n7 more\n0.7%   Solanaceae\n0.6%   Lamiales\n0.6%   campanulids\n0.6%   Triticeae\n0.6%   PACMAD clade\n1%   Sar\n16 more\nViruses\nRiboviria\nOrtho...viraePisuviricota\n8%   Severe acute respiratory syndrome coronavirus 2N...a\n6 more\nRevtra...icetesO...s\n1%   Human immunode\n\"ciency virus 1\nBacteria\nenvi...ples\n5%   uncultured bacterium\nPseudomonadota\n0.9%\n   Gammaproteobacteria\n8 more\nTerr...roup\n0.8%\n   Bacillota\n10 more26 more\nLength (log10)\n1.0\n2.4\n3.8\n5.1\n6.5\n7.9\nA\nB\nC\nFigure S1. Recentrifuge’s scripts allow the characterization and interactive exploration of the content of the nt database during the building process of the nt database for\nCentrifuge. (A) Recentrifuge’srcf interactive plot showing the content of the nt database for the number of sequences and the logarithm of their length in the color scale. (B)\nRecentrifuge’srefafilt output showing the distribution of the sequence lengths in the nt database. (C) Recentrifuge’srcf interactive plot displaying new bacterial taxa not\npresent in the previous seasonally compiled database showing the number of sequences and the logarithm of their length in the color scale.\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint \n\nMartí et al. | 9\nin millions (106)\nFigure S2. Example of inter-kingdom contamination network plot generated during the process of building the nt database for Centrifuge. The contamination was detected\nbetween the five default \"kingdoms\" as defined in Conterminator software, with the width and color scale of the arrows related to the number of sequences involved. The rate\nof contamination detected in the database (non-redundant sequences) before contamination removal was significant: almost half a percentage point.\nmammal_equal_mix\nmammal_log_mix\nnon−mammal_mix\nprotozoa_mix\nzymo_gut_microbiome\nzymo_micro_comm_II\nATCC_20_strain\natcc_gut_spike\natcc_mycobiome\natcc_oral_spike\nATCC_path_mix\natcc_skin_spike\nJun22_decSep22_decJan23_decJan23_stdJan23_filt\nApr23_decon\nApr23_std Jun22_decSep22_decJan23_decJan23_stdJan23_filt\nApr23_decon\nApr23_std Jun22_decSep22_decJan23_decJan23_stdJan23_filt\nApr23_decon\nApr23_std Jun22_decSep22_decJan23_decJan23_stdJan23_filt\nApr23_decon\nApr23_std Jun22_decSep22_decJan23_decJan23_stdJan23_filt\nApr23_decon\nApr23_std Jun22_decSep22_decJan23_decJan23_stdJan23_filt\nApr23_decon\nApr23_std\n0.0\n0.2\n0.4\n0.6\n0.0\n0.2\n0.4\n0.6\ndatabase\nBC distance\nsample\nATCC_20_strain\natcc_gut_spike\natcc_mycobiome\natcc_oral_spike\nATCC_path_mix\natcc_skin_spike\nmammal_equal_mix\nmammal_log_mix\nnon−mammal_mix\nprotozoa_mix\nzymo_gut_microbiome\nzymo_micro_comm_II\nA\nB\n0.00 0.25 0.50 0.75 1.00\n1\n10\n100\n1000\nabundance threshold\nno. of false positives\nmhl\nmhl15\nmhl22\nmhl30\nmhl40\nmhl50\n0.00 0.25 0.50 0.75 1.00\n1\n10\n100\nabundance threshold\nno. of false positives\ntax_rank\nphylum\nclass\norder\nfamily\ngenus\nspecies\nFigure S3. Additional comparative assessment of gDNA reference controls. A) Bray Curtis dissimilarity distance between database versions across all gDNA reference control\nsamples. B) Number of false positive assignments between MHL values and taxonomic rank across abundance thresholds.\n(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for this preprintthis version posted June 14, 2024. ; https://doi.org/10.1101/2024.06.12.598617doi: bioRxiv preprint","source_license":"Public-Domain","license_restricted":false}