Centriflaken: An automated data analysis pipeline for assembly and in silico analyses of foodborne pathogens from metagenomic samples Authors: Kranti Konganti, Julie A Kase, Narjol Gonzalez-Escalona, Mark Eppinger, Mark Eppinger, Mark Eppinger, Mark Eppinger Published: 2025 ABSTRACT AbstractRapid and comprehensive analysis of metagenomic data from samples associated with foodborne outbreaks is of critical importance in food safety. Equally important is the need for automated analysis pipelines that allow the rapid and effective construction of metagenomic assembled genomes (MAGs) to enable bacterial source-tracking from metagenomic data. Here, we present centriflaken, an automated precision metagenomics pipeline for detecting and characterizing Shiga toxin-producing Escherichia coli (STEC) from metagenomic data. centriflaken streamlines the process of generating metagenome-assembled genomes (MAGs) and conducting in silico analyses of STECs, significantly reducing the time and manual effort required for comprehensive pathogen profiling. centriflaken was validated using Oxford Nanopore long-read sequencing data from agricultural water enrichments, successfully reproducing results from our previous study that involved multiple manual bioinformatics steps (Maguire et al., 2021). The tool’s efficacy was further demonstrated through its application to ZymoBIOMICS microbial community standards and 21 additional irrigation water samples, completing STEC precision metagenomics analyses in less than 7 hours per sample. centriflaken’s versatility allows for the analysis of user-defined taxa beyond STEC, including other foodborne pathogens like Listeria monocytogenes or Salmonella. The pipeline generates comprehensive summary plots and tables, accessible through a MultiQC HTML report. Designed for portability, centriflaken packages all software dependencies within containers and virtual environments. This open-source tool is available on GitHub under the MIT license (https://github.com/CFSAN-Biostatistics/centriflaken), offering a powerful resource for rapid, automated pathogen detection and characterization in food safety applications. Introduction As farm-to-table food trends increase in popularity, we have seen an increase in produce related outbreaks of foodborne illness. Shiga toxin-producing E. coli (STEC) are a major source of foodborne illness [1–4]. In fact, O157:H7 accounts for 70% of E. coli related foodborne outbreaks [5–10]. Recent outbreaks underscore the ongoing threat of STEC contamination. For instance, in October 2024, a multi-state outbreak caused by a O121:H19 STEC strain linked to organic carrots resulted in 48 illnesses, 20 hospitalizations, and 1 death, highlighting the critical need for rapid and accurate STEC detection and characterization methods (https://www.cdc.gov/ecoli/outbreaks/e-coli-o121.html). Due to the ubiquitous nature and the varying degrees of pathogenicity of E. coli, its mere presence does not equate to concerns to public health, unlike with bacteria such as Salmonella or Listeria monocytogenes. Therefore, the detection of E. coli presence alone is insufficient to assess health risk. Moreover, there are over 400 described STEC serotypes that express one or more of stx1 genes and stx2 genes [11,12] and among those, around 100 serotypes have been described as causing illness in humans, with hemolytic uremic syndrome (HUS) being deadly [13–18]. However, serotype and stx gene presence and variant type are not enough to predict STEC risk for humans. STEC must also have other virulence genes linked to attachment and colonization of the intestine such as intimin (eae gene), AggR (aggR gene), auto-agglutinating adhesin Saa (saa gene), chromosome located T3SS effectors encoded by genes such as tir, espA, espB, espK, among others, and other virulence genes located in plasmids (toxB, hexA, etpD, katP) [19–23]. In our previous study, we established the use of nanopore sequencing to determine the presence and full characterization of STECs in enrichments of agricultural water [24]. We determined that the limit for obtaining a complete, fragmented assembly for STECs by nanopore sequencing of DNA extracted from enriched agricultural water was 105 CFU/ml [24]. There are many advantages of using nanopore sequencing for STEC identification and characterization: affordability, potential to multiplex, portability, less manual labor required from personnel, use of low DNA concentration, longer reads than other available technologies (>100 kb) and the analysis could also be done in real-time (using their metagenomics workflow available at EPI2ME labs). A previous pilot study produced preliminary evidence that MinION sequencing of agricultural water using the ligation kit has the potential to be used for rapid microbiome determination in the field with optimal results for water quality surveillance [25]. Agricultural water has been implicated as a source of contamination for produce resulting in foodborne illness and outbreaks [26–31]. Current FDA protocols for the detection and isolation of STECs require multiple rounds of selective plating and whole genome sequencing (WGS) of a single isolate. On-site field testing is increasingly becoming a priority to decrease the time for detection of pathogenic microbes and for the fast implementation of corrective measures. While the most abundant species in the microbiome is likely to fluctuate seasonally, they may be an indicator to changing populations and importantly may serve to monitor deviations in microbial water quality. However, any of these studies or experiments will generate a large amount of data. Analyzing nanopore data or any other high-throughput data is a very complex procedure and usually requires many steps to generate the final reportable results creating a bottleneck. Metagenomic workflows are usually compute intense and need to be analyzed on High-Performance Compute Clusters (HPCs). In the case of data generated from metagenomic samples, tools are needed to detect the presence of STEC and comprehensively assess their virulence potential by subtyping the stx genes, determining serotype, identifying attachment genes such as eae and saa and detecting other virulence-related genes. A precision metagenomics approach is not meant to replace the cultural method [32], but it is a way to detect and identify a STEC strain earlier and evaluate its potential risk to the public. Improving STEC genome characterization methods has significant implications for public health and food safety. Rapid, accurate identification and genome characterization of STECs can lead to faster response times in outbreak situations, more targeted recalls, and ultimately, fewer cases of foodborne illness. Moreover, enhanced detection capabilities could inform more precise and effective preventive controls in food production and processing, aligning with FDA’s risk-based approach to food safety. The existing protocol or pipeline in our FDA laboratory uses many sequential manual steps, described in maguire et al. (2021) [24]. Briefly, the steps are divided into: 1) classify reads using WIMP workflow available at EPI2me, 2) a python script to extract or bin reads that matched the requested taxa (i.e., E. coli), 3) Assemble the extracted reads using flye assembler [33], 4) Classify the contigs with Kraken 2 [34] and extract contigs matching requested taxa, 5) perform all in silico analysis on those contigs (MLST, AMR, serotyping, and virulotyping), and 6) aggregate all data into a final report and conduct a risk assessment for the sample. Several other workflows for metagenomic analyzes have been published and reviewed by van damme et al. 2021 [35]. These authors described the creation of a very useful metagenomic tool that allows for the analyses of metagenomic data using both short and long reads in a hybrid approach which allows for the reconstruction of more complete and more accurate metagenome-assembled genomes (MAGs) [35]. While these approaches have strengths, none of them allows for a precision metagenomic analysis (taxa specific) and in silico virulotyping, serotype, and AMR prediction for STECs in a single workflow. Here we present centriflaken to address this gap. centriflaken is an integrated suite of automated data analysis workflows enabled by Nextflow [36]. This pipeline processes metagenomic data to generate metagenome-assembled genomes (MAGs) and performs comprehensive in silico analysis for Shiga toxin-producing Escherichia coli (STEC), following the approach described by Maguire et al. (2021). Centriflaken streamlines the entire process, from raw sequencing data to final results, and provides user-friendly output in the form of summary plots and tables, accessible through a MultiQC HTML report generated automatically by the pipeline. The development of centriflaken aligns closely with FDA’s New Era of Smarter Food Safety initiative, which emphasizes the use of new technologies and approaches to create a safer, more digital, and traceable food system. By enabling faster, more comprehensive analysis of metagenomic data, centriflaken supports FDA’s goals of strengthening root cause analyses, improving predictive analytics, and responding more rapidly to outbreaks. Materials and methods Centriflaken pipeline A high-level overview of centriflaken workflow is presented in Fig 1. The pipeline can be run on any UNIX based machine. The main analysis starts by taking UNIX input paths to the fastq_pass reads folder generated by Oxford Nanopore. It then filters out any reads less than 4000 bp and then performs taxonomic classification and extracts all E. coli only reads using centrifuge [37] followed by taxon specific assembly using flye assembler v2.9 [33]. The assembled contigs are reclassified to identify the E. coli taxon using kraken2 [34] and extract the E. coli contigs to produce a final Metagenomically assembled Genome (MAG). The final MAG is interrogated (in the case of STEC) against several specific databases. Serotypefinder [14] is used to serotype the MAGs, ABRicate (https://github.com/tseemann/abricate) is run to identify known AMR genes using various databases (AMRfinder plus [38], MEGARes 2.0 [39], RESFINDER [40], ARG-ANNOT [41]) and their results are shown in a final MultiQC report. All the analysis steps are automated using Nextflow and the pipeline runs in parallel for all samples. A brief overview of the centriflaken pipeline. The pipeline starts by taking as input a UNIX path to FASTQ files after which multiple read files belonging to the same sample are merged followed by excluding reads whose length is less than 4000 bp. Then, reads are extracted belonging to a taxon of interest using Centrifuge following which the target assembly is performed using flye. Kraken2 is now used to extract contigs binned as the taxa of interest and these contigs are used for further downstream analysis such as subtyping, virulotyping, AMR gene finding, etc. The outputs from all these downstream processes are used to generate a summary report using MultiQC. The centriflaken GitHub repository provides guidance on acquiring the necessary Centrifuge and Kraken databases. For Centrifuge, links to pre-built indices are available, while for Kraken, step-by-step instructions for database construction are provided. Detailed information can be found in the repository’s README file at: https://github.com/CFSAN-Biostatistics/centriflaken/blob/main/README.md#databases. It is recommended to run centriflaken using conda software virtual environments for users who do not have access to HPC cluster with singularity. The pipeline can also be run using docker or singularity containers by switch the -profile flag when invoking the pipeline. Detailed instructions on setup and running the pipeline on data from Maguire et al. 2021 [24] can be found at https://github.com/CFSAN-Biostatistics/centriflaken/blob/main/readme/centriflaken.md. centriflaken is also made available via Galaxy user interface via the FDA HFP GalaxyTrakr instance (https://galaxytrakr.org). The previously published protocol (https://www.protocols.io/view/centriflaken-an-automated-data-analysis-pipeline-f-kxygxzdbwv8j/v5) can be used to run analysis on GalaxyTrakr. While centriflaken is designed to work with both Illumina short reads and Oxford Nanopore long reads, this publication specifically focuses on its application to long-read sequencing data generated by Oxford Nanopore technologies. An example of centriflaken’s performance using Illumina short reads can be found in S4 MultiQC HTML Report in S4 File. Validation of centriflaken and accuracy assessment Centriflaken was initially tested and validated using data from Maguire et. al. 2021 [24]. The presence of the complete genome and synteny of the completely closed genomes on the final assemblies was checked using the whole genome alignment tool available in the CLC Genomics workbench software v24.0 (QIAGEN, Redwood City, USA). Further testing was conducted with data generated from two ZymoBIOMICS Microbial Community Standards (ZymoBIOMICS Microbial Community Standard II – log distribution – catalog number D6311, and ZymoBIOMICS Microbial Community DNA Standard – catalog number D6306) (ZymoBIOMICS, Irvine, CA), and can be found on GenBank using the accession numbers in S1 Table. The centriflaken final testing was done using 21 overnight enriched samples for STEC collected in two different seasons during 2020 and 2021 (sample collection and processing detailed below). Overnight enrichments of agricultural waters Water samples were collected from various sites, and no permits were required because a member of the study team was affiliated with the relevant authority. All samples were prepared according to the established protocol: dx.doi.org/10.17504/protocols.io.j8nlk8nydl5r/v1. Briefly, water samples (225 ml) from the Southwestern US were enriched by adding an equal volume of 2 × modified Buffered Peptone water with pyruvate (mBPWp; NEOGEN, Lansing, MI, USA) as described in USFDA Bacteriological Analytical Manual (BAM) Chapter 4A [32], with 6.0 mL of acriflavin-cefsulodin-vancomycin (ACV) supplement added after five hours of static incubation at 37°C ± 1°C. After the addition of the ACV supplement, enrichments were incubated at 42°C ± 1°C for an additional 16 ± 3 hours. DNA extraction Genomic DNA from 1 ml of each of the 21 enriched agricultural water sample was extracted using the Maxwell RSC Cultured Cell DNA Kit with a Maxwell RSC Instrument (Promega Corporation, Madison, WI, USA) according to the manufacturer’s instructions DNA concentration was determined by a Qubit 4 Fluorometer (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. STEC qPCR detection The presence of STEC was determined by qPCR as described in Chapter 4A of the FDA BAM detecting stx1, stx2, and wzy [32]. Briefly, the DNA retrieved from the 21 samples were diluted 1:10 in nuclease-free water and 2µl was added to 28µl master mix containing 0.25µM stx1 and stx2 primers, 0.3µM wzy primers, 0.2µM stx1 probe, 0.15µM stx2 and wzy probes, 1X Internal Positive Control Mix (Cat: 4308323, Applied Biosystems), 1X Express qPCR Supermix Universal Taq (Cat: 11785200, Invitrogen), and ROX passive dye. All primers and probes (S2 Table) employed in this study were purchased from IDT (Coralville, IA, USA). Nanopore sequencing DNA recovered from the 21 enriched AW samples was sequenced using a GridION nanopore sequencer (Oxford Nanopore Technologies, Oxford, UK). The sequencing libraries were prepared using the Genomic DNA by Ligation kit (SQK-LSK109) and run in FLO-MIN106 (R9.4.1) flow cells, according to the manufacturer’s instructions for 72 hours (Oxford Nanopore Technologies). The runs were live base called using either Guppy v3.2.10 or 4.2.3 included in the MinKNOW v19.12.6 and v20.10.6 software (Oxford Nanopore Technologies, Oxford, UK), respectively, with the fast-calling model. The two ZymoBIOMICS Microbial Community Standards (ZymoBIOMICS Microbial Community Standard II – log distribution – catalog number D6311, and ZymoBIOMICS Microbial Community DNA Standard – catalog number D6306) (ZymoBIOMICS, Irvine, CA), were also sequenced as described above using 1 ug of DNA as the initial amount for preparing the library. Results Software pipeline overview The data analysis steps in centriflaken are automated using the Nextflow workflow manager [36], which provides advantages such as process parallelization, sample provenance, retry on failure, tool dependency solutions using containers and conda software virtual environments among many others. Successful execution of the workflow produces output for each process in its own output folder named after the process name. The workflow design also ensures reproducibility of results, and it is easy to share HTML brief reports for further assessment. Due to the inherent advantages of using Nextflow, centriflaken can be run on-premises HPC clusters or in the cloud. Initial validation with known samples The entire centriflaken workflow was initially executed and validated on 7 samples from the Bioproject PRJNA639799 (available at NCBI) using conda software virtual environments. This metagenomic study provided nanopore-based data for each water enriched sample spiked with different STEC levels to determine detection and classification of STECs using nanopore sequencing [24]. These nanopore read dataset provided differential coverage for the spiked STEC strain, and the workflow runtime for all seven samples was around 20 hours. The entire process when performed manually took almost 24 hours for a single sample [24]. centriflaken reduced the analysis time from 24 hours per sample to approximately 3 hours per sample, representing an 87.5% reduction in processing time. Both the manual workflow and the automated Nextflow version were evaluated on the Reedling HPC Cluster, which uses SLURM for job scheduling with each compute node containing 48 CPU cores and 256 GB of memory. Each of the individual steps of the centriflaken workflow were executed in batch mode (i.e., each command was written sequentially as a bash script). In contrast the Nextflow version automatically parallelizes the entire sequence of pipeline steps. centriflaken recovered STEC genomes with genome completeness of at least 85% when the concentration of spiked STEC was above 106 CFU/ml (Table 1 and Fig 2). For comparison, the number of total contigs per STEC spiked water sample when performing the step by step analysis nanopore data workflow [24] was a little bit higher than when using centriflaken, but a similar number of contigs containing the STEC spiked genome was recovered (with minimal differences observed only at the two recoverable lowest inoculation levels), confirming the equivalency of both analysis workflows (Table 1). However, centriflaken has a step that extracts the contigs matching the requested taxa and therefore the analysis is only performed in those contigs, accelerating and making the analysis more precise (see workflow on Fig 1). The pipeline analysis also showed the same virulence profile for each sample to what was reported previously in Maguire et al 2021 (Table 2, S1 MultiQC HTML Report in S1 File), demonstrating the accuracy and reproducibility of the centriflaken workflow. This high level of genome completeness at concentrations above 106 CFU/ml demonstrates centriflaken’s sensitivity and accuracy in detecting and characterizing STEC in complex metagenomic samples, a crucial capability for food safety applications. In accordance with standard FDA food safety protocols for STECs, an enrichment step is performed prior to sequencing to amplify target pathogen concentrations from potentially low initial levels to the detectable range (≥106 CFU/ml). Table 1In silico virulence factor analysis of metagenomic assembled genomes (MAGs) obtained using centriflaken for the Maguire et al. (2021) dataset. Showing the agreement of the results obtained by both workflows (manually vs. centriflaken).SampleEDL933 Inoculation Level(CFU/ml)Serotypestx type eae Maguire et al 2021Contig No.centriflakenContig No.Flow cell number/sample fastqEDL 933 genome coverage type (genome and plasmid)a(genomeand plasmid)a%Water0O9––312FAL009400Water+Ecoli17.3 x 108O157:H71agamma-144 (1+)8 (1+)FAL 11217100Water+Ecoli27.3 x 107O157:H71agamma-140 (1+)11 (1+)FAL01664100Water+Ecoli37.3 x 106O157:H7/ O91agamma-141 (8+)30 (9+)FAL11341100Water+Ecoli4b7.3 x 105O157:H7/ O91agamma-192 (63+)60 (59+)FAL0090885Water+Ecoli57.3 x 104O9––284FAL008700Water+Ecoli67.3 x 103O9––242FAL016620aIn parenthesis, the number of contigs that contain the entire chromosome, a plus sign (+) indicates the presence of a contig that matched the EDL933 circular plasmid.bfragmented genome assembly limit. In silico virulence factor analysis of metagenomic assembled genomes (MAGs) obtained using centriflaken for the Maguire et al. (2021) dataset. Showing the agreement of the results obtained by both workflows (manually vs. centriflaken). aIn parenthesis, the number of contigs that contain the entire chromosome, a plus sign (+) indicates the presence of a contig that matched the EDL933 circular plasmid. bfragmented genome assembly limit. Comparative analysis of EDL933 genome assemblies obtained by centriflaken during the initial validation with known samples. This figure compares the EDL933 genome of the strain used in Maguire et al. (2021) with assemblies obtained using centriflaken for the same samples at various EDL933 enrichment spiking levels. The analysis demonstrates the recovery of the E. coli O157:H7 MAG (Metagenome-Assembled Genome) in either completely closed or fragmented forms (see Table 1). Each horizontal track represents EDL933-matching contigs extracted from a sample. Homologous segments across genomes are indicated by the same color and connected. Sequence coordinates in base pairs are shown on respective scales. The reference genome, labeled as EDL933_2 and indicated with a star (*), serves as the basis for comparison. Table 2Comparative results of the virulotyping of the same ONT data generated (Maguire et al 2021) [24] by both workflows (manually vs. centriflaken).E. coli Maguire 2021E. coli centriflakenVirulence GeneEDL933referenceWatera123456Water123456 astA +–++++–––++++–– ehxA +–++++–––++++–– espA +–++++–––++++–– espB +–++++–––++++–– espF +–++++–––++++–– espJ +–++++–––++++–– espK +–++++–––++++–– espP +–++++–––++++–– tccP +–++++–––++++–– etpD +–++++–––++++–– gad +–++++–––++++–– iha +–++++–––++++–– iss +–++++–––++++–– nleA +–++++–––++++–– nleB +–++++–––++++–– nleC +–++++–––++++–– tir +–++++–––++++–– katP +–++++–––++++–– pssA +–++++–––++++–– air +–++++–––++++–– toxB +–++++–––++++––aSamples: Water (0 CFU/ml STEC EDL933), 1 (7.3 x 108 CFU/ml), 2 (7.3 x 107 CFU/ml), 3 (7.3 x 106 CFU/ml), 4 (7.3 x 105 CFU/ml), 5(7.3 x 104 CFU/ml), and 6 (7.3 x 103 CFU/ml). Comparative results of the virulotyping of the same ONT data generated (Maguire et al 2021) [24] by both workflows (manually vs. centriflaken). aSamples: Water (0 CFU/ml STEC EDL933), 1 (7.3 x 108 CFU/ml), 2 (7.3 x 107 CFU/ml), 3 (7.3 x 106 CFU/ml), 4 (7.3 x 105 CFU/ml), 5(7.3 x 104 CFU/ml), and 6 (7.3 x 103 CFU/ml). Performance with ZymoBIOMICS microbial communities The centriflaken workflow was further tested on samples from a previous study using the ZymoBIOMICS microbial community DNA Bioproject PRJNA751542, Biosample SAMN20520875 [25] and two additional samples (ZymoBIOMICS Microbial Community DNA Standard and ZymoBIOMICS Microbial Community Standard II – log distribution) (NCBI accession provided here). Both communities consist of a mixture of 10 microbes belonging to different genera and species but at different proportions. ZymoBIOMICS microbial community composition is at a 12% concentration each bacterium and 2% for the two yeasts, while ZymoBIOMICS Microbial Community Standard II composition is quite different. ZymoBIOMICS Microbial Community Standard II composition is as follows: of Listeria monocytogenes – 89.1%, Pseudomonas aeruginosa – 8.9%, Bacillus subtilis – 0.89%, Saccharomyces cerevisiae – 0.89%, Escherichia coli – 0.089%, Salmonella enterica – 0.089%, Lactobacillus fermentum – 0.0089%, Enterococcus faecalis – 0.00089%, Cryptococcus neoformans – 0.00089%, and Staphylococcus aureus – 0.000089%. Only the ZymoBIOMICS microbial community DNA samples contained enough E. coli (108 CFU per 1 ug) to be within the limits of genome assembly (>106 CFU/ 1 ug) as establish previously for ONT metagenomic samples using the LSK-109 library preparation kit [24] (s2 multiQC HTML Report in S2 File). centriflaken identified the E. coli presence in the sample belonging to serotype O33:H32 and carried 4 known AMR genes using the NCBI AMRfinder plus tool (blaEC, mdtM, emrD, and acrF) [38]. The ZymoBIOMICS Microbial Community Standard II – log distribution contained ~ 105 E. coli CFU per 1 ug, well below the assembly limit for the technique. That sample still returned a signal for E. coli (159 unique taxa matching reads) but no E. coli MAGs were recovered and therefore the correct identity as well as the virulence or serotype characteristics of the E. coli strain could not be determined (s2 multiQC HTML Report in S2 File). These results highlight centriflaken’s ability to accurately identify and characterize E. coli strains even in diverse microbial communities, while also demonstrating its limitations at lower bacterial concentrations. This information is vital for understanding the tool’s applicability in various real-world scenarios. Analysis of unknown agricultural water After validating the centriflaken workflow with known samples, the next step was to assess its performance with ONT data obtained after sequencing unknown overnight enriched agricultural water samples suspected of containing STECs. For that purpose, we tested the performance on 21 samples obtained after overnight enrichment of agricultural water. qPCR screening of these 21 enriched samples confirmed the presence of stx (17/21) and O157 wzy (12/21) genes. The results are summarized in Table 3. Four samples tested negative for either stx or O157 wzy genes and were used as negative control for evaluating the centriflaken pipeline. Table 3qPCR results for stx and wzy genes for the 21 enriched samples.SampleqPCR O157 wzy (Cq value)qPCR stx gene typeSTEC concentration (CFU/ml)aFAO96374+ (25.7)1, 25.45E + 05FAL11425–––FAL11145+ (27.3)1, 21.87E + 05FAL11151+ (24.3)21.48E + 06FAO72169+ (27.5)1, 21.58E + 05FAL00958–––FAL11351+ (22.3)1, 28.87E + 05FAL05306+ (23.8)1, 21.98E + 06FAL16059–1, 27.22E + 05FAQ34034+ (25.9)1, 24.87E + 05FAO71993+ (27.6)1, 21.485E + 05FAL11417+ (27.5)1, 21.485E + 05FAO97359+ (22.5)24.99E + 06FAL17459–––FAQ33890+ (24.3)1, 21.42E + 06FAQ33923b+ (22.6)1, 24.72E + 06FAL11387–––FAL01556–1, 27.93E + 05FAL11305–1, 25.55E + 05FAL11384b–1, 22.85E + 06FAL01198b–1, 27.26E + 06aThe CFU/ml were calculated based on either the O157 wzy or stx genes qPCR results.bstx genes were also detected in the assembled nanopore data.(+) positive. (-) negative. qPCR results for stx and wzy genes for the 21 enriched samples. aThe CFU/ml were calculated based on either the O157 wzy or stx genes qPCR results. bstx genes were also detected in the assembled nanopore data. (+) positive. (-) negative. The entire precision metagenomics STEC analysis of the 21 samples was completed in less than 7 hours using centriflaken (S3 MultiQC HTML Report in S3 File). Fig 3 shows the full metagenomic analysis workflow, starting from sample collection to final reporting, including the initial culturing step. centriflaken was run on a Reedling HPC Cluster, which uses SLURM for job scheduling with each compute node containing 48 CPU cores and 256 GB of memory. The total reads per experiment ranged from 1.3 to 2.8 million reads, of which 11–40% matched E. coli, with median read sizes between 4,252–6,177 bp (Table 4). The number of total E. coli contigs per sample ranged from 178 to 1,069. Workflow of the complete precision metagenomic analysis. The figure shows the sequential steps from sample collection and overnight culturing to DNA extraction, qPCR, library preparation, Oxford Nanopore sequencing, and data analysis using the centriflaken pipeline, ending with the generation of the final report. This overview summarizes the rapid, end-to-end process used for the detection and characterization of STEC in agricultural water samples. Table 4Sequencing statistics, serotypes, stx genes and eae genes, identified per sample.Flow cell IDTotal readsmedian read length (bp)E. coli readsSerotypes identified(O types)stx type eae contig No.E. colistrainsFAO963741,663,8834252422,626O10,O103,O109,O110,O113,O123,O128ac,O15,O16,O162,O166,O18,O185,O33,O6,O7,O73,O8,O82,O83,O84,O88,O9a––59623FAL114251,961,1204687211,801O153,O170,O178,O185,O8,O81,O82,O93––1788FAL111451,492,2834577358,148O10,O110,O123,O128ac,O139,O148,O15,O162,O18ac,O20,O21,O3,O33,O55,O6,O7,O73,O82,O89,O9––37820FAL111511,279,8665452287,970O10,O110,O146,O150,O156,O162,O166,O175,O18ac,O20,O33,O38,O6,O8,O82,O84–+38916FAO721691,993,1744262460,423O1,O106,O110,O125ab,O139,O148,O168,O170,O174,O185,O18ac,O27,O33O4,O75,O79,O8,O81,O82,O92,O9a–+55521FAL009582,048,3454748374,847O101,O103,O106,O11,O113,O117,O13/O129,O136,O140,O153,O156,O166,O179,O18,O2,O29,O33,O61,O7,O74,O8,O81,O84–+64723FAL113512,130,8714628447,483O106,O113,O139,O140,O142,O18ac,O20,O33,O36,O43,O48,O63,O7,O82,O88,O9––66216FAL053062,268,6984807499,114O1,O104,O106,O109,O110,O123,O124,O139,O140,O147,O15,O150,O156,O162:H6,O166,O20,O28ac/O42,O33,O4,O55,O6,O7,O8,O81,O88––57825FAL160592,074,8225150466,835O12,O125ab,O128ac,O140,O142,O162,O166,O23,O26,O51,O6,O62,O7,O70,O74,O8,O84,O88,O9––52919FAQ340342,128,4404917542,752O10,O100,O101,O103,O104,O128ab,O132,O138,O139,O15,O171,O178,O18,O20,O33,O36,O6,O7,O73,O8,O82,O83,O88,O89,O93––53525FAO719932,263,6184928563,641O105,O106,O123,O125ab,O132,O142,O150,O156,O163,O166,O170,O179,O181,O22,O23,O3,O33,O4,O46,O51,O52,O59,O6,O7,O8,O86,O88,O9––64728FAL114171,912,5155366535,504O102,O12,O125ab,O139,O140,O142,O146,O153,O154,O16,O162,O166,O167,O18ac,O20,O21,O22,O23,O26,O43,O46,O51,O6,O62,O7,O73,O8,O82,O86,O88,O9a––65531FAO973592,778,5546177630,732O10,O105,O110,O112ab,O132,O146,O147,O150,O157,O16,O162,O17/O77,O170,O18,O20,O28ac,O33,O45,O53,O55,O6,O8,O82,O93,O9a–+65925FAL174592,460,1215671509,245O106,O113,O150,O156,O174,O175,O178,O18ac,O21,O22,O28ab,O29,O6,O7,O8,O82,O93–+49717FAQ338902,222,4475289575,614O1,O102,O103,O121,O123,O139,O140,O142,O15,O150,O153,O162,O163,O166,O171,O18ac,O20,O28ac/O42,O33,O36,O4,O45,O48,O5,O51,O6,O7,O73,O8,O81,O82,O86,O88,O89,O9–+81535FAQ339232,364,5904759766,127O10,O102:H21,O104,O105,O113,O123,O134,O139,O140,O15,O150,O162,O166,O18,O20,O33,O36,O45,O6,O7,O71,O73,O74,O8,O82,O882+69526FAL113872,569,3555312565,258O104,O106,O113,O126,O132,O139,O140,O141,O150,O154,O163,O166,O170,O179,O18ac,O20,O22,O29,O36,O45,O48,O52,O7,O8,O81,O83,O85,O93–+61428FAL015562,459,1805256639,387O101,O104,O106,O12,O128ac,O13/O129,O142,O153,O18ac,O21,O22,O23,O29,O3,O52,O6,O62,O7,O8,O9,O93,O99––54222FAL113052,719,6715451734,311O101,O12,O140,O142,O146,O15,O20,O23,O29,O3,O43,O52,O6,O62,O7,O8,O88,O9––51918FAL113842,638,8545043976,376O10,O100,O101,O102,O110,O114,O12,O123,O124,O128ac,O130,O134,O139,O140,O142,O146,O148,O150,O153,O154,O16,O163,O166,O172,O178,O18ac,O2,O21,O22,O23,O24,O29,O3,O36,O43,O54,O6,O62,O65,O69,O7,O73,O8,O80,O84,O88,O91, 2–94947FAL011982,503,39053911,001,356O10,O101,O104,O112ac,O113,O116,O117,O12,O121,O126,O128ac,O130,O139,O140,O142,O146,O148,O150,O153,O16,O166,O169,O171,O175,O178,O18ac,O20,O21,O22,O23,O28ac/O42,O29,O45,O5,O51,O6,O62,O7,O73,O75,O8,O82,O83,O84,O86,O8,O9,O9a2–106948 Sequencing statistics, serotypes, stx genes and eae genes, identified per sample. Every sample contained more than one different E. coli strain, with varying degrees of diversity. If the predicted serotypes identified among the many MAGs obtained per sample is used as a measure of E. coli diversity, the number of E. coli strains per sample ranged from eight to forty-eight (Table 4). The identification of multiple E. coli strains in each sample showcases centriflaken’s capacity to detect and differentiate various strains in complex environmental samples, a key advantage for comprehensive food safety assessments. There were only 3 samples that were identified as positive for the stx genes but one of them (FAL11384) lacked any adhesion genes used in our search (Tables 3 and 4). Sample FAL01198 was positive for subA and FAQ33923 detected the eae gene allele epsilon-7. On the other hand, 8 samples returned a signal for the eae gene, and one sample contained saa. The virulotyping showed the presence of several relevant virulence genes such as T3SS (eae, espA, espB, espD, espF, espI, espJ, and tir) and other plasmid borne virulence relevant genes (espP and exhA) (Table 5), however a fully closed STEC genome was not recovered from any sample. While a fully closed STEC genome was not recovered, the detection of numerous virulence genes demonstrates centriflaken’s potential for comprehensive pathogen profiling, providing valuable insights for risk assessment in food safety contexts. AMR profiling showed the presence of a very diverse set of antimicrobial resistance genes per sample, from 2 to up to 20 genes, representing resistance to 10 antibiotic classes and multidrug efflux genes, underscores the potential of centriflaken for comprehensive pathogen characterization beyond just virulence factors (S3 Table). Table 5STEC virulotyping for 21 enriched agricultural waters by centriflaken.Sample eae saa suba exhA espA espB espD espF espI espJ espP tccP tir FAO96374–––+–––––––––FAL11425–––––––––––––FAL11145–––+–––––––––FAL11151+–––+++–––––+FAO72169++–+++++–++–+FAL00958+––++++––+––+FAL11351–––++++–––––+FAL05306–––+–––––––––FAL16059–––+–––––––––FAQ34034–––+–––––––––FAO71993–––+–––––––––FAL11417–––+–––––––––FAO97359+–––+++–––––+FAL17459+––+++++––––+FAQ33890+––+++++–+–++FAQ33923+––++++++–––+FAL11387+––––––––+–––FAL01556–––+–––––––––FAL11305–––+–––––––––FAL11384–––+–––––––––FAL01198––++––––––+–– STEC virulotyping for 21 enriched agricultural waters by centriflaken. Discussion Importance of rapid and automated metagenomic analysis for food safety Given the critical role of irrigation water in food production, rapid and accurate detection of STECs in these waters is essential but especially during outbreak investigations. Traditional detection methods, such as qPCR followed by extensive selective plating and whole genome sequencing (WGS), are labor-intensive and may take up to two weeks to confirm the correct identification of an isolate. Fast metagenomic analysis is crucial for accelerating this timeline. By integrating qPCR with long read metagenomic sequencing of the pre-enrichment step, definitive detection and virulence characterization of STECs can be achieved within 3–4 days. While microbiological confirmation is still required, this approach potentially enables faster implementation of corrective actions since overall turnaround time is reduced by almost one week. Centriflaken: Addressing bioinformatic bottlenecks However, the bioinformatic analysis of sequencing data [24] remains complex, involving many tools and steps to generate a complete STEC analysis. This process often requires up to 24 hours per sample -even when those samples analyses are run in parallel. To address this bottleneck, a workflow named centriflaken was developed to automate the entire process and to reduce the analysis time. centriflaken enables taxa-specific precision metagenomic analysis, in silico virulotyping, serotyping, and AMR prediction for STECs in a single, reproducible workflow. centriflaken consolidates the different steps of a precision metagenomic approach into a scalable, highly reproducible, and easy-to-install workflow using Nextflow [36]. It outputs final reports, summary plots and tables through a MultiQC HTML report for user friendly review as part of the pipeline. By automating and streamlining the analysis process, centriflaken not only saves time but also reduces the potential for human error, enhancing the reliability and reproducibility of STEC detection in complex environmental samples. centriflaken uses many tools to provide a comprehensive analysis of ONT metagenomics samples. The associated tools were chosen based on the steps used for the analysis of ONT samples previously published [24] and by benchmark performance, e.g., taxonomic classification and binning [37], E. coli reads extraction [24], assembly [33], taxonomic classification [34], and final in silico profiling [14,38]. Built with a modular Nextflow DSL2 design, centriflaken is easily updated to include improved tools in future revisions or new ONT chemistries,such as the recent shift from R9.4.1 to R10.4.1. Performance benchmarking Centriflaken was first evaluated using previous analyzed data under Bioproject PRJNA639799 [24], and two ZymoBIOMICS microbial community datasets. For STEC-spiked samples, centriflaken achieved 100% genome completeness at concentrations >10⁷ CFU/ml and 85% at 10⁶ CFU/ml—mirroring the results from manual, stepwise analyses [24]. Serotyping, virulence gene detection, and AMR profiling were also consistent with previously reported outputs. In the ZymoBIOMICS microbial community dataset, centriflaken accurately identified the E. coli strain present in the community as serotype O33:H32 and detected four AMR-related genes, including blaEC, mdtM, emrD, and acrF [42]. In contrast, the E. coli genome was not recovered from the ZymoBIOMICS Microbial Community Standard II (Log Distribution), as its concentration was below the ~ 10⁶ CFU threshold necessary for effective assembly [24]. In spiked agricultural water samples, centriflaken recovered STEC genomes with ≥85% completeness at ≥10⁶ CFU/ml (Table 1 and Fig 2). Although manual analysis produced slightly more contigs overall [24], the number of contigs corresponding to the STEC genome was equivalent, confirming the equivalency of both analysis workflows (Table 1). centriflaken also includes a step that extracts only the contigs belonging to the target taxa, streamlining the process and improving accuracy (Fig 1). The virulence profiles identified were consistent with prior results from Maguire et al. (2021) [24], confirming the accuracy and reproducibility of the workflow (Table 2, S1 MultiQC HTML Report in S1 File). Real-world application and limitations The performance of the centriflaken workflow was further evaluated using 21 real-life enriched agricultural water samples. The results showed that each sample contained multiple E. coli strains—ranging from 2 to 13—based on predicted serotype profiles. The presence of so many strains per sample, combined with their high concentrations as measured by qPCR (~10⁶ CFU/ml) (Table 3), made it difficult to generate fully closed MAGs. This finding highlights a key limitation of the methodology: recovering and characterizing STEC genomes in silico is challenging when samples are highly complex and contain multiple E. coli strains, especially when these strains outcompete the STEC target. Despite this challenge, three samples were identified as containing stx genes and adhesion genes such as saa or eae. A representative case was sample FAQ33923, which contained 26 distinct E. coli serotypes, including O104, O113, and O45—three serotypes commonly associated with human infections. Several virulence genes, including stx2 and eae, were detected. The eae gene matched an Escherichia albertii intimin allele (eae epsilon-7/Xi allele, GenBank accession FJ609833.1), while stx2 was located on a separate contig, suggesting that these genes originated from different strains within the same enrichment. This observation underscores the limitations of metagenomic reconstruction in complex microbial communities, where fragmented assemblies and strain diversity can obscure the true genomic context of virulence determinants. An additional eight samples produced E. coli MAGs that harbored eae genes. The eae subtypes identified included: gamma-2 (in two samples), epsilon-7/Xi (in four samples), and beta-1 (in one sample). Regarding AMR prevalence in those 21 samples, between 2 and 20 genes were detected per sample. These genes potentially confer resistance to 10 antibiotic classes, including multidrug efflux systems. While these limitations highlight areas for future improvement, they also underscore the complexity of real-world samples and the importance of centriflaken’s ability to detect multiple strains simultaneously. Future directions and potential impact Centriflaken continues to evolve. The modular structure of the workflow setup of centriflaken using Nextflow allows further refinements as soon as future developments in ONT sequencing or metagenomics arise, as observed recently with the introduction of newer chemistries (R10.4.1) [43–51]. Nextflow-based parallelization enhances computational efficiency. With R10.4.1’s improved basecalling accuracy [43–48], there is growing potential to recover complete STEC genomes directly from complex samples—enabling phylogenetic comparisons to public genomes via the NCBI Pathogen Detection database. (https://www.ncbi.nlm.nih.gov/pathogens/). ONT long reads further facilitate genome closure by resolving repetitive regions, as previously demonstrated with STEC EDL933 O157:H7 [24]. This is fully leveraged in centriflaken’s reassembly step, which focuses analysis on user-defined taxa to reduce computational load and maximize accuracy by only using the addressed taxa through a precision metagenomic approach as previously described [24]. centriflaken is publicly available at: https://github.com/CFSAN-Biostatstics/centriflaken. Centriflaken supports any user-defined taxon, including Listeria monocytogenes and Salmonella, expanding its utility for foodborne pathogen surveillance. All software dependencies are containerized for portability and reproducibility. The development of centriflaken as a useful tool in combination with metagenomic analysis is an important step in evaluation of a variety of samples associated with fresh produce. As a result, metagenomic analysis can now be completed within hours rather than weeks—providing precise detection and characterization of STEC and other pathogens in tested samples. While initially focused on STEC, centriflaken’s flexible design allows for adaptation to other foodborne pathogens. This versatility positions it as a valuable tool for comprehensive food safety monitoring, potentially revolutionizing how we approach pathogen detection and outbreak investigation. Looking to the future Centriflaken is currently available to Human Foods Program users via the HFP HPC cluster and GalaxyTrakr: https://www.protocols.io/view/centriflaken-an-automated-data-analysis-pipeline-f-kxygxzdbwv8j/v5. As ONT long-read metagenomics continues to mature, tools like centriflaken could play a transformative role in culture-independent pathogen detection and source tracking. Its portability makes it suitable for point-of-care testing and field deployment in food and water testing scenarios. Future studies should focus on optimizing centriflaken for even lower bacterial concentrations, integrating it with other emerging technologies like machine learning for predictive analytics, and validating its performance across a wider range of food and environmental matrices. Conclusion Centriflaken significantly accelerates and streamlines STEC detection in environmental samples. It provides a scalable, automated alternative to labor-intensive traditional workflows, delivering fast, reproducible, and accurate results. With continued enhancements and integration of newer ONT chemistries (e.g., R10.4.1), centriflaken may support real-time surveillance of foodborne pathogens directly from complex samples and shorten response times in outbreak scenarios. Furthermore, ongoing improvements and adaptations will potentially enhance its utility for food safety monitoring. In conclusion, centriflaken represents a substantial advancement in metagenomic analysis for food safety applications. By dramatically reducing analysis time and complexity while maintaining high accuracy, it has the potential to transform STEC detection and characterization, ultimately contributing to more rapid and effective food safety interventions. Supporting information S1 MultiQC HTML Report. centriflaken MultiQC HTML Report for Maguire et al. (2021) samples. (HTML) S2 MultiQC HTML Report. centriflaken’s output for the ZymoBIOMICS microbial community DNA samples. (HTML) S3 MultiQC HTML Report. centriflaken’s output for the 21 agricultural water samples analyzed. (HTML) S4 MultiQC HTML Report. centriflaken’s performance using Illumina short reads. (HTML) Raw ONT data generated for each individual sample in this study (ZymoBiomics and the 21 agricultural water samples). (XLSX) qpCR primers and probes used in this study. (XLSX) Centriflaken AMR profiling of the 21 enriched agricultural waters. (XLSX) Data Availability All relevant data are within the manuscript and its Supporting Information files. Funding Statement This work was supported by FDA Human Foods Program Intramural Funds (no award number assigned) and FDA Office of the Commissioner Chief Scientist Challenge Grant (Award Number: 2021-1464) to Narjol Gonzalez-Escalona. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. References 11 Sep 2025 Dear Dr. González-Escalona, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Oct 26 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols . We look forward to receiving your revised manuscript. Kind regards, Mark Eppinger Academic Editor PLOS ONE Journal Requirements: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. Please note that PLOS One has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse. 3. In your Methods section, please provide additional information regarding the permits you obtained for the work. Please ensure you have included the full name of the authority that approved the field site access and, if no permits were required, a brief statement explaining why. 4. We note that the grant information you provided in the ‘Funding Information’ and ‘Financial Disclosure’ sections do not match. When you resubmit, please ensure that you provide the correct grant numbers for the awards you received for your study in the ‘Funding Information’ section. 5. Thank you for stating the following in the Acknowledgments Section of your manuscript: The study was supported by funding from the Chief Scientist-Challenge Grants Proposal #2021-1464 and the FDA Foods Program Intramural Funds. We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form. Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows: The author(s) received no specific funding for this work. Please include your amended statements within your cover letter; we will change the online submission form on your behalf. 6. Please amend your list of authors on the manuscript to ensure that each author is linked to an affiliation. Authors’ affiliations should reflect the institution where the work was done (if authors moved subsequently, you can also list the new affiliation stating “current affiliation:….” as necessary). 7. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information. 8. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. 9. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice. [Note: HTML markup is below. Please do not edit.] Reviewers' comments: Reviewer's Responses to Questions Comments to the Author 1. Is the manuscript technically sound, and do the data support the conclusions? Reviewer #1: Yes ********** 2. Has the statistical analysis been performed appropriately and rigorously? -->?> Reviewer #1: N/A ********** 3. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy Reviewer #1: Yes ********** 4. Is the manuscript presented in an intelligible fashion and written in standard English??> Reviewer #1: Yes ********** Reviewer #1: This manuscript describes the development of a stand-alone open-source tool (centriflaken) for the automated detection of Shiga toxin-containing Escherichia coli (STEC) in metagenomic samples. Previously, the authors published the analysis pipeline, but each subsequential step in the pipeline had to be manually started. This manuscript compares and contrasts the previous pipeline with the new stand-alone version. The ability to parallelize the new version resulted in the analysis completing in about 1/3 the time it took the previous version and maintained the same accuracy. Another added feature is the use of singularity to package all the software with their dependencies which will help end-users install and run centriflaken. This new tool has the ability to help identify STEC in metagenomic samples taken from water if the concentration meets the minimum detection limit. Questions/comments Page 5, line 121. change “bysubtyping” to “by subtyping”. Page 7, line 152. When starting a sentence, the “c” in centriflaken should be capitalized. Also, page 9, line 193, the “c” in centriflaken is capitalized. My guess is that it should be lowercase. Page 8, lines 176. The sentence “The pipeline took approximately 6 hours to finish….” Is a result in the Material and Methods section. There needs to be more information about the run including number of reads, total number of bases, was it run in parallel, what processors were used, and other spec of the compute system being used. Page 9, lines 193-195. Why isn’t more information given about using Illumina short reads with centriflaken? This is asking the reader to take your word that it works instead of showing the data. This is the perfect opportunity to show the process for using centriflaken with short read technologies. Page 9, lines 205-207. Please include more information about the samples used in the study. Did you re-extract DNA from the water sample, was the water sample frozen, was the DNA frozen from a previous extraction or did you use the reads generated from a previous Nanopore run? Page 9, line 212. I am not sure what the “X” means in the sentence. Page 12, line 263. Change “provides” to “provided”. Page 12, lines 265-268. Please provide the more detail about the computer systems used to run centriflaken and the previous manual version. Page 12, lines 273-275. The sentences says that there were the “same” number of contigs between the previous version and centriflaken. While they are the same for two inoculation levels, they are similar for the lower inoculation level samples. Table 1. There is a superscript “c” at the end of the Maguire et al 2021 column. It probably should be a superscript “a”. Page 13, lines 280-283. What is the concentration of pathogens found in food outbreak vectors? Is there 106 CFU/ml? When thinking of complex samples to extract DNA for metagenomic samples from, water would not be on the top of my list. How would this be used with other preharvest sample types. What happens when DNA is extracted from food stuffs like meat products or leafy greens where eukaryotic DNA can make up most of the extracted DNA. The usefulness of this tool is very apparent. I am just wondering how logistical it is to use with samples that have low pathogen numbers and is diluted in DNA from other organisms. Page 14, line 319. Again, please provide some type of context to the computers systems used. Tables 3 and 4. How would you interpret the results from sample FAQ33923 because it has stx2, eae, tir exhA, espA, espB, espD, espF and espI but there multiple E. coli serotype associated with sample. There are several serotypes identified in the serotype analysis that have been isolated from human infections including O104, 0113, and O45. Do you have enough sequence data to type the eae gene? I am afraid that the results from this sample could be also interpreted that eae and stx are located in isolates from different serotypes. Tables. Please list the samples in the Tables in the same order. Page 16, lines 252-254. qPCR was mentioned in the Materials and Methods but not in the results. It would be good to have a Figure showing the complete fast metagenomic analysis procedure including culturing. Supplemental Table 3. The genes under the macrolide column needs to be italicized. ********** what does this mean? ). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy Reviewer #1: No ********** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org . Please note that Supporting Information files do not need this step. 19 Nov 2025 Reviewer's Responses to Questions Comments to the Author Reviewer #1: This manuscript describes the development of a stand-alone open-source tool (centriflaken) for the automated detection of Shiga toxin-containing Escherichia coli (STEC) in metagenomic samples. Previously, the authors published the analysis pipeline, but each subsequential step in the pipeline had to be manually started. This manuscript compares and contrasts the previous pipeline with the new stand-alone version. The ability to parallelize the new version resulted in the analysis completing in about 1/3 the time it took the previous version and maintained the same accuracy. Another added feature is the use of singularity to package all the software with their dependencies which will help end-users install and run centriflaken. This new tool has the ability to help identify STEC in metagenomic samples taken from water if the concentration meets the minimum detection limit. - Thank you for your thoughtful and comprehensive review of our manuscript. We appreciate your recognition of the key improvements in centriflaken, particularly the automation capabilities, parallelization benefits, and the singularity packaging for enhanced usability. Your acknowledgment of the tool's potential application for STEC detection in water samples is especially valuable, and we're grateful for your constructive feedback on this work. Page 5, line 121. change “bysubtyping” to “by subtyping”. -done. Page 7, line 152. When starting a sentence, the “c” in centriflaken should be capitalized. Also, page 9, line 193, the “c” in centriflaken is capitalized. My guess is that it should be lowercase. - Agreed and done. Page 8, lines 176. The sentence “The pipeline took approximately 6 hours to finish….” Is a result in the Material and Methods section. There needs to be more information about the run including number of reads, total number of bases, was it run in parallel, what processors were used, and other spec of the compute system being used. - I agreed. We decided to eliminate this sentence from the materials and methods section and move it to the results section, as described below in response to the Page 12, lines 265-268 comment. Page 9, lines 193-195. Why isn’t more information given about using Illumina short reads with centriflaken? This is asking the reader to take your word that it works instead of showing the data. This is the perfect opportunity to show the process for using centriflaken with short read technologies. - We appreciate the reviewer's valid concern regarding the limited information provided about using Illumina short reads with centriflaken. The reviewer is correct that additional data and demonstration would strengthen this manuscript. We acknowledge that the current presentation asks readers to accept our assertion without sufficient supporting evidence. While this is not the primary focus of this manuscript, we do have preliminary data on centriflaken's performance with Illumina short reads available on protocols.io. These results represent part of a separate, more comprehensive study using Illumina MiSeq reads that is currently in preparation for publication. This forthcoming study will provide detailed validation data, performance metrics, and comparative analyses specifically focused on short-read sequencing applications with centriflaken. To address the reviewer's concern in the current manuscript, we will add a supplementary MultiQC file showing results for 29 short-read samples with diverse levels of STECs and the following sentence: “An example of centriflaken's performance using Illumina short reads can be found in Supplementary MultiQC HTML Report 4.” Page 9, lines 205-207. Please include more information about the samples used in the study. Did you re-extract DNA from the water sample, was the water sample frozen, was the DNA frozen from a previous extraction or did you use the reads generated from a previous Nanopore run? - We thank the reviewer for requesting clarification about our sample preparation methodology. The 21 samples were collected in 2020 and 2021, and both sample processing and DNA extraction were performed immediately upon collection. Nanopore sequencing runs were conducted as soon as DNA was obtained. Due to the COVID-19 pandemic, our research focus was temporarily diverted to other studies, and we recently re-analyzed the existing sequencing data with centriflaken for this publication. To provide complete transparency regarding our experimental design, we have modified the sentence to include "(sample collection and processing detailed below)" as suggested by the reviewer. This clarifies that we used fresh samples with immediate processing and sequencing, rather than re-extracted DNA, frozen samples, or previously generated sequencing reads. Page 9, line 212. I am not sure what the “X” means in the sentence. - Thank you for your observation. That was a typo. We have eliminated the X. Page 12, line 263. Change “provides” to “provided”. - Agreed and done. Page 12, lines 265-268. Please provide the more detail about the computer systems used to run centriflaken and the previous manual version. - The reduction of time and computational resources mentioned in the paper are with respect to running them on our HPC Cluster. Both the individual steps that are part of the centriflaken workflow and the automated nextflow workflow were run on Reedling HPC Cluster which uses SLURM for job scheduling with each compute node containing 48 CPU cores with 256 GB of memory. Each of the individual steps of the centriflaken workflow were executed in batch mode (i.e., each command was written sequentially as a bash script). In contrast the Nextflow version automatically parallelizes the entire sequence of pipeline steps. We have added sentence addressing this: “Both the manual workflow and the automated Nextflow version were evaluated on the Reedling HPC Cluster, which uses SLURM for job scheduling with each compute node containing 48 CPU cores and 256 GB of memory. Each of the individual steps of the centriflaken workflow were executed in batch mode (i.e., each command was written sequentially as a bash script). In contrast the Nextflow version automatically parallelizes the entire sequence of pipeline steps.” Page 12, lines 273-275. The sentences says that there were the “same” number of contigs between the previous version and centriflaken. While they are the same for two inoculation levels, they are similar for the lower inoculation level samples. - Agreed and modified the sentence to read as: “but a similar number of contigs containing the STEC spiked genome was recovered (with minimal differences observed only at the two recoverable lowest inoculation levels), confirming the equivalency of both analysis workflows” Table 1. There is a superscript “c” at the end of the Maguire et al 2021 column. It probably should be a superscript “a”. - That is correct. Thank you for the correction. Done. Page 13, lines 280-283. What is the concentration of pathogens found in food outbreak vectors? Is there 106 CFU/ml? When thinking of complex samples to extract DNA for metagenomic samples from, water would not be on the top of my list. How would this be used with other preharvest sample types. What happens when DNA is extracted from food stuffs like meat products or leafy greens where eukaryotic DNA can make up most of the extracted DNA. The usefulness of this tool is very apparent. I am just wondering how logistical it is to use with samples that have low pathogen numbers and is diluted in DNA from other organisms. - The reviewer raises an important point about pathogen concentrations in real-world food samples. In FDA food safety protocols for STEC detection and isolation, an enrichment step is always performed prior to analysis. This enrichment process increases the target pathogen numbers from potentially very low initial concentrations (which may be below detection limits) to higher, detectable levels typically in the range we tested (10^6 CFU/ml or higher). Therefore, the concentrations we evaluated in this study reflect the post-enrichment levels that would be encountered in actual FDA food safety workflows, not the initial pathogen loads in raw food samples. The enrichment step is a standard and essential component of food pathogen detection protocols, ensuring that even samples with initially low pathogen numbers can be effectively analyzed. This makes centriflaken directly applicable to current FDA food safety practices, as it would be used on enriched samples where pathogen concentrations have been amplified to detectable levels. That is why the initial STEC qPCR of the enrichment is a very crucial step in the entire workflow, because that will provide the concentration of the target in the sample and the feasibility of closing or getting a fragmented STEC genome by long read sequencing. We have added this sentence after that statement: “In accordance with standard FDA food safety protocols for STECs, an enrichment step is performed prior to sequencing to amplify target pathogen concentrations from potentially low initial levels to the detectable range (≥106 CFU/ml).” Page 14, line 319. Again, please provide some type of context to the computers systems used. - Agreed and done. We have added the following sentence: “centriflaken was run on a Reedling HPC Cluster, which uses SLURM for job scheduling with each compute node containing 48 CPU cores and 256 GB of memory.” Tables 3 and 4. How would you interpret the results from sample FAQ33923 because it has stx2, eae, tir exhA, espA, espB, espD, espF and espI but there multiple E. coli serotype associated with sample. There are several serotypes identified in the serotype analysis that have been isolated from human infections including O104, 0113, and O45. Do you have enough sequence data to type the eae gene? I am afraid that the results from this sample could be also interpreted that eae and stx are located in isolates from different serotypes. - We thank the reviewer for this insightful question. We have added a new explanation to the Discussion section to clarify the interpretation of sample FAQ33923. Specifically, we now note that this sample contained 26 distinct E. coli serotypes, including O104, O113, and O45—three serotypes commonly associated with human infections. Several virulence genes, including stx2 and eae, were detected. The eae gene showed 99% identity to an Escherichia albertii intimin allele (epsilon-7/Xi allele, GenBank accession FJ609833.1), while stx2 was located on a separate contig, suggesting that these genes originated from different strains within the same enrichment. Other virulence factors (espA, espB, espD, and espF) also matched E. albertii, supporting the interpretation that eae and its associated genes derive from a non-STEC background. Although stx2 was clearly present, the high degree of genomic fragmentation and diversity of E. coli serotypes prevented conclusive linkage between stx2 and eae. This case exemplifies the challenge of analyzing highly mixed metagenomic samples where multiple closely related E. coli strains coexist, underscoring the limitations of metagenomic reconstruction in resolving the genomic context of virulence determinants. Tables. Please list the samples in the Tables in the same order. - Agreed and done. We thank the reviewer for their excellent suggestion. Page 16, lines 252-254. qPCR was mentioned in the Materials and Methods but not in the results. - We thank the reviewer for this helpful comment. We have added a description of the qPCR results in the Results section as follows: “qPCR screening of these 21 enriched samples confirmed the presence of stx (17/21) and O157 wzy (12/21) genes. The results are summarized in Table 3. Four samples tested negative for either stx or O157 wzy genes and were used as negative control for evaluating the centriflaken pipeline”. The results are summarized in a new Table (Table 3), and four samples that tested negative for either stx or wzy were used as negative controls for evaluating the centriflaken pipeline. It would be good to have a Figure showing the complete fast metagenomic analysis procedure including culturing. - We agree with the reviewer’s suggestion. We have added a new figure (Figure 3) illustrating the complete fast metagenomic analysis workflow, starting from sample collection and culturing to final reporting. This figure provides a clear visual overview of the end-to-end process described in the manuscript. Supplemental Table 3. The genes under the macrolide column needs to be italicized. - Agreed and done. Submitted filename: Response to Reviewers PONE-D-25-38731.docx 25 Nov 2025 centriflaken: an automated data analysis pipeline for assembly and in silico analyses of foodborne pathogens from metagenomic samples PONE-D-25-38731R1 Dear Dr. González-Escalona, We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements. Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication. An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support . If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. Kind regards, Mark Eppinger Academic Editor PLOS ONE Additional Editor Comments (optional): Reviewers' comments: PONE-D-25-38731R1 PLOS One Dear Dr. González-Escalona, I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team. At this stage, our production department will prepare your paper for publication. This includes ensuring the following: * All references, tables, and figures are properly cited * All relevant supporting information is included in the manuscript submission, * There are no issues that prevent the paper from being properly typeset You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps. Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org. You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing. If we can help with anything else, please email us at customercare@plos.org. Thank you for submitting your work to PLOS ONE and supporting open access. Kind regards, PLOS ONE Editorial Office Staff on behalf of Dr. Mark Eppinger Academic Editor PLOS One Supplementary Materials S1 MultiQC HTML Report. centriflaken MultiQC HTML Report for Maguire et al. (2021) samples. (HTML) S2 MultiQC HTML Report. centriflaken’s output for the ZymoBIOMICS microbial community DNA samples. (HTML) S3 MultiQC HTML Report. centriflaken’s output for the 21 agricultural water samples analyzed. (HTML) S4 MultiQC HTML Report. centriflaken’s performance using Illumina short reads. (HTML) Raw ONT data generated for each individual sample in this study (ZymoBiomics and the 21 agricultural water samples). (XLSX) qpCR primers and probes used in this study. (XLSX) Centriflaken AMR profiling of the 21 enriched agricultural waters. (XLSX) Submitted filename: Response to Reviewers PONE-D-25-38731.docx Data Availability Statement All relevant data are within the manuscript and its Supporting Information files.