- Open Access
Benchmarking of two bioinformatic workflows for the analysis of whole-genome sequenced Staphylococcus aureus collected from patients with suspected sepsis
BMC Infectious Diseases volume 23, Article number: 39 (2023)
The rapidly growing area of sequencing technologies, and more specifically bacterial whole-genome sequencing, could offer applications in clinical microbiology, including species identification of bacteria, prediction of genetic antibiotic susceptibility and virulence genes simultaneously. To accomplish the aforementioned points, the commercial cloud-based platform, 1928 platform (1928 Diagnostics, Gothenburg, Sweden) was benchmarked against an in-house developed bioinformatic pipeline as well as to reference methods in the clinical laboratory.
Whole-genome sequencing data retrieved from 264 Staphylococcus aureus isolates using the Illumina HiSeq X next-generation sequencing technology was used. The S. aureus isolates were collected during a prospective observational study of community-onset severe sepsis and septic shock in adults at Skaraborg Hospital, in the western region of Sweden. The collected isolates were characterized according to accredited laboratory methods i.e., species identification by MALDI-TOF MS analysis and phenotypic antibiotic susceptibility testing (AST) by following the EUCAST guidelines. Concordance between laboratory methods and bioinformatic tools, as well as concordance between the bioinformatic tools was assessed by calculating the percent of agreement.
There was an overall high agreement between predicted genotypic AST and phenotypic AST results, 98.0% (989/1006, 95% CI 97.3–99.0). Nevertheless, the 1928 platform delivered predicted genotypic AST results with lower very major error rates but somewhat higher major error rates compared to the in-house pipeline. There were differences in processing times i.e., minutes versus hours, where the 1928 platform delivered the results faster. Furthermore, the bioinformatic workflows showed overall 99.4% (1267/1275, 95% CI 98.7–99.7) agreement in genetic prediction of the virulence gene characteristics and overall 97.9% (231/236, 95% CI 95.0–99.2%) agreement in predicting the sequence types (ST) of the S. aureus isolates.
Altogether, the benchmarking disclosed that both bioinformatic workflows are able to deliver results with high accuracy aiding diagnostics of severe infections caused by S. aureus. It also illustrates the need of international agreement on quality control and metrics to facilitate standardization of analytical approaches for whole-genome sequencing based predictions.
Infectious diseases caused by bacteria are one of the leading causes of human mortality and morbidity throughout the world, being responsible for several million deaths each year . One of the most common pathways to death following an infection is sepsis, which arises when the body’s systemic response to an infection injures its own tissues and organs. It can lead to multiple organ dysfunction, shock and death, especially if not recognized early and treated promptly. Every hour of delayed appropriate antibiotic therapy increases mortality in septic shock by 5–10% [2, 3]. Early identification of patients having bacterial sepsis along with timely determination of causative bacteria and antibiotic resistance profiles can alter current practices for therapeutic management, reduce over-prescription of antibiotics and associated adverse outcomes [4, 5]. The rapidly increasing area of next-generation sequencing (NGS) technologies and more specifically bacterial whole-genome sequencing (WGS) could offer several applications in clinical microbiology including accurate and earlier species identification of bacteria, prediction of antimicrobial resistance and virulence genes . However, bacterial WGS has seen slow integration into routine microbiological diagnostics because of the lack of a platform that can translate WGS data into clinical practice. Furthermore, WGS workflows are required to be standardized when considering the clinical diagnostic application . To comply with the aforementioned points the commercial cloud-based platform, 1928 platform (1928 Diagnostics, Gothenburg, Sweden) was benchmarked against an in-house developed bioinformatic pipeline (INH) as well as to reference methods in the clinical laboratory. In the present study WGS data were retrieved from 264 Staphylococcus aureus isolates, using the Illumina HiSeq X next-generation sequencing technology. The S. aureus isolates were collected and characterized according to accredited laboratory methods during a prospective observational study of community onset of severe sepsis and septic shock in adults at Skaraborg Hospital, in the western region of Sweden . The outcomes from the clinical laboratory methods, i.e., species identification by MALDI-TOF MS (Bruker) analysis and phenotypic antibiotic susceptibility testing (AST) following the EUCAST guidelines, were viewed as reference results, the true results, for the comparison with the genome-based computationally predicted output from the bioinformatic analyses of the WGS data. In addition, the virulence gene predictions obtained from the 1928 platform were compared to those obtained from the INH. Results for multi-locus sequence typing (MLST) were also investigated. Such application of WGS bioinformatics methods aiding in S. aureus diagnostics has also been addressed in other studies [9,10,11]. A recent study, analyzing WGS data from blood culture isolates of S. aureus using Next Gen Diagnostic software (Mountain View, California, USA) and the 1928 platform reported slightly high very major error (VME) and major error (ME) rates for the 1928 platform . One VME is defined as a resistant phenotype with genetic predicted susceptible genotype, also known as false negatives. One ME is defined as a susceptible phenotype with genetic predicted resistant genotype, also known as false positives . Since false negatives can have consequences for treatment of infection [14, 15] the present study compared the numbers of VME and ME, but also the VME and ME rates retrieved from the investigated bioinformatic workflows. Furthermore, the 1928 platform has been used in two studies focused on S. argenteus [16, 17] and in another study investigating the regional epidemiology and susceptibility patterns of methicillin resistant S. aureus (MRSA) isolates identified in Stockholm County, Sweden . Lastly, a recent publication looking into the biodiversity of clinical Klebsiella spp. isolates collected from patients with suspected community-onset sepsis, Sweden, included bioinformatic analyses by the 1928 platform . The present study aimed at to further evaluate the performance of the 1928 platform in clinical routine for in silico species identification, antibiotic susceptibility testing, virulence and sequence typing of S. aureus. Since rapid extraction of clinically relevant genomic information will be essential for the adoption of WGS for infection control and public health, the processing time of the bioinformatic workflows was also compared.
From September 2011 to June 2012, a prospective observational study of community-onset severe sepsis and septic shock in adults was conducted at Skaraborg Hospital, a secondary hospital with 640 beds, in the western region of Sweden . The study was approved by the Regional Ethical Review Board of Gothenburg (376–11). As the present study only focused on bacterial isolates recovered from cultures included in the routine patient care, individual informed consent is deemed unnecessary according to national regulations (2003:460). Approximately 1,800 bacterial isolates were recovered at the clinical microbiology laboratory, Unilabs, Sweden, from the patients enrolled in the study. These isolates were cryopreserved at the time of recovery by transferring colonial material to Microbank™ vials (Pro-Lab Diagnostics, Ontario, Canada) stored at − 80 °C. For the present study, isolates recovered from 212 patients and identified as S. aureus (n = 272) with routine microbiological methods based on cultures followed by MALDI-TOF MS (DB-4110) were selected. Nevertheless, five isolates could not be recovered after freezing. In all, 267 isolates were prepared for DNA extraction and WGS (Fig. 1).
Reference method for species identification
In line with the hospital’s policy, blood cultures were drawn from each patient before the initiation of intravenous antibiotic treatment. For patients with sepsis of unknown origin, samples from urine and respiratory tract were cultured whenever possible. Other samples were collected at the discretion of the treating physician. Microbiological culturing was performed as previously described . Definite species identification of the collected isolates was performed by MALDI-TOF MS on a Microflex LT mass spectrometer (Bruker Daltonics, Leipzig, Germany) using BioTyper software v2.0 using default parameter settings as described elsewhere [20, 21]. Spectral scores above 2.0 were used as a cut-off for correct identification. At the time of the study, the Bruker microorganism database MBT Compass Library DB-4110 (Bruker Daltonics, Germany) released in April 2011 was used.
Reference method for antibiotic susceptibility testing
Antibiotic susceptibility was determined by accredited laboratory methods using the disc diffusion method on Mueller Hinton media according to European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines (www.eucast.org). Antibiotic susceptibility test (AST) results retrieved from the identified S. aureus were included, hereafter referred to as phenotypic AST. Resistant AST result for isoxazolyl penicillin and cefoxitin was followed by detection of mecA by PCR to confirm the isolate as a methicillin resistant S. aureus (MRSA). Phenotypic AST results reported in this study are limited to the set of antibiotics included in the 1928 platform (Table 1).
Whole-genome sequencing of S. aureus—Illumina HiSeq
Genomic DNA was extracted at Unilabs, Skövde, using the MagNA Pure 96 DNA and Viral NA Small Volume kit (Roche Diagnostics, Switzerland) with the Pathogen Universal 200 protocol on a MagNA Pure 96 instrument (Roche Diagnostics, Switzerland). DNA concentration was measured using the Qubit dsDNA HS assay kit (Thermo Fisher Scientific, USA) on a Qubit 3.0 (Thermo Fisher Scientific, USA) and NanoDrop spectrophotometer, respectively (Thermo Fisher Scientific, USA). DNA extracts from 267 S. aureus isolates were transported to Clinical Genomics, SciLifeLab, Solna, Sweden, where the WGS was performed. During the sample preparation Nextera XT DNA sample preparation guide (Illumina, USA), was followed. Measurement of double-stranded DNA concentration was achieved with broad- and low range assay kits on a Qubit 2.0 (Thermo Fisher Scientific, USA). Library preparation was performed according to the Nextera XT guidelines (Illumina, USA). Fragment analyses of the PCR libraries on a Bioanalyzer (Agilent technologies, USA) was done to obtain abundances and average length of fragments for each sample. The Illumina HiSeq 2500 platform was used for the NGS. The output files consisted of compressed FASTQ-files (.gz) containing sequencing data that could be downloaded for further analysis.
The INH consisted of established bioinformatic tools as illustrated in Fig. 2. In more detail, primary quality control of the FASTQ files was performed using the FastQC software (v.0.11.8) . One isolate was removed from the dataset before trimming as it showed low number of reads (< 500,000). Trimmomatic (v.0.36) was used for adapter removal and quality trimming with a sliding window of size 4 and a minimum quality of 20 . In addition, the first 12 bases were trimmed by the HEADCROP argument, and reads with a length shorter than 30 bp were removed. FASTQ files were then assembled into contigs using SPAdes (v.3.13.1) . The quality of the assembled contigs were evaluated using the QUAST (v.5.0.2)  with default settings. In addition to determining assembly metrics, the length of each assembly was manually compared to the genome size of a reference genome obtained from NCBI. The reference genome used was Staphylococcus aureus subsp. aureus, NCTC 8325 (GenBank accession number NC_007795.1) with a genome size of 2.8 Mbp. If an assembly was not considered good, the median coverage was also calculated using R v.3.5 . Two genome assemblies had a median coverage < 2.5 reads per base and were excluded from further analysis. The assembled contigs in FASTA format were annotated by tools available in the Center for Genomic Epidemiology (CGE) (https://www.genomicepidemiology.org/) i.e., ResFinder v.3.0 , VirulenceFinder v2.0 , MLST 2.0  and the JSpeciesWS (http://jspecies.ribohost.com/jspeciesws/) (Fig. 2). Species identification was achieved by calculating the pairwise average nucleotide identity (ANI) based on BLAST + (ANIb) in JSpeciesWS  using Staphylococcus aureus subsp. aureus NCTC 8325 as the reference genome. An ANI threshold of 96% or greater was considered to delineate species boundaries as a threshold of 96% correlates well to DNA-DNA hybridization [28, 29] (Fig. 2). The presence of antibiotic resistance genes was predicted using CGE ResFinder v.3.0 with default settings for threshold ID (90%) and minimum length 60% . Susceptibility was conferred by the absence of resistance genes and resistance was conferred by the presence of resistance genes. Presence (P) or absence (A) of virulence genes were predicted by CGE VirulenceFinder v2.0  with default settings for threshold ID (90%) and minimum length 60%. MLST analysis was performed using CGE MLST 2.0  with Staphylococcus aureus as selected configuration for all isolates. This MLST scheme consists of alleles from the following seven loci arcC, aroE, glp, gmk, pta, tpi, and yqiL . For analysis with 1928, the FASTQ files were uploaded to its cloud-based platform (1928 Diagnostics, Sweden) for inferred antibiotic susceptibility based on genotype resistance markers, which are genes and mutations known to contribute to antibiotic resistance, hereby and later referred to as predicted genotypic antibiotic susceptibility. Susceptibility was conferred by the absence of genotype resistance markers and resistance was conferred by the presence of genotype resistance markers. The platform also predicts acquired virulence genes, type of mobile genetic SCCmec element and sequence type (Table 1), hereby and later referred to as predicted genotypic presence (P) or absence (A) of different virulence genes and ST, respectively. After uploading the FASTQ files, the files underwent an initial quality control were reads were trimmed or entirely discarded according to the platform’s internal thresholds i.e., sequencing depth/coverage higher than 30 × to perform the analysis. During species identification, analysis depth/coverage of 11-29X was allowed. Species identification, gene and mutation detection for the other analyses were performed by assembly free kmer-based methods (Table 2). Raw pair-end fastq.gz files were uploaded to the 1928 platform during June and July 2019. This platform has not been further updated during the access period as confirmed by communication with 1928 Diagnostics, Sweden.
Benchmarking of the bioinformatic workflows
Species identification and AST results retrieved from the reference method in the clinical laboratory were compared to genetic predicted species identification and AST by both bioinformatic workflows through examining the degree of agreement between these results (Table 2). Nine antibiotics and/or antibiotic classes were included in the benchmarking (Table 3). A very major error (VME) was defined as a resistant phenotype with genetic predicted susceptible genotype, a major error (ME) was defined as a susceptible phenotype with genetic predicted resistant genotype . Acquired virulence genes, presence (P) or absence (A), and STs retrieved from the bioinformatic workflows were compared by examining the degree of agreement between the genetically predicted results.
Statistical analyses and calculations were performed using R v.4.0.3 . Concordance between conventional microbiological methods and bioinformatic analysis tools was assessed by calculating the percent of agreement. Concordance between bioinformatic analysis tools was assessed by calculating the percent of agreement. The Agresti-Coull method was used  for the construction of 95% CI for percent of agreement between methods. Jupyter notebook v.6.0.3  in Anaconda v. 2–2.4.0  was used for pre-processing data.
The current study was based on 264 isolates identified as S. aureus according to the reference method using MALDI-TOF MS analysis (Fig. 1). Using the 1928 platform, nine FASTQ files did not pass the internal quality control levels, since sequencing depth/coverage had to be higher than 30 × to perform the analysis (1928 Diagnostics, Sweden). Allowing depth/coverage of 11-29X for the species identification of these nine isolates, the 1928 platform showed 99.2% (262/264, 95% CI 97.1–99.9) agreement with the reference method. For the two discrepant results the 1928 platform predicted one of the isolates as Staphylococcus epidermidis (SA 310) whereas the second isolate (SA 1413) was predicted as non-staphylococcus spp. The INH also showed 99.2% (262/264, 95% CI 97.1–99.9) agreement to the reference method. Among the two discrepant results, the first isolate (SA 310) was also predicted as S. epidermidis, whereas the second isolate (SA 1413) was predicted as Staphylococcus argenteus. The two samples identified as S. epidermidis and S. argenteus were excluded from the rest of the study, since this study focused on S. aureus. Likewise, the FASTQ files that did not pass the internal quality control levels in the 1928 platform (1928 Diagnostics, Sweden) were excluded from further analysis. The upcoming benchmarking of virulence gene characterization, sequence type and antibiotic susceptibility included the remaining 255 S. aureus isolates.
Antibiotic susceptibility test
In all, the EUCAST testing for the clinical S. aureus isolates generated 1006 phenotypic AST results (Table 4). The AST showed 2.5% (25/1006) isolates to be phenotypically resistant and the highest percentage of resistance was noticed for ciprofloxacin and fusidic acid (Table 4). Three out of 244 clinical isolates were phenotypically resistant to isoxazolyl penicillin (Table 4), being reported as cefoxitin resistant, which were further confirmed by in-house PCR as mecA gene positive and reported as methicillin resistant S. aureus (MRSA).
In order to study the capacity of the bioinformatic workflows to translate WGS into clinical practice within the area of AST, the degree of agreement between the predicted genotypic AST to the phenotypic AST result was studied (Table 5). There was an overall high agreement 98.0% (989/1006, 95% CI 97.3–99.0). Among the phenotypic AST results reported as susceptible 99.8% (979/981, 95% CI 99.2–99.9) of the results were concordant, while the phenotypic AST results reported as resistant were less concordant 60.0% (15/25, 95% CI 40.7–76.5) (Table 5). The degree of agreement between the predicted genotypic AST to the phenotypic AST result from 1928 and INH using ResFinder (Fig. 2) showed 99.0% (996/1006, 95% CI 98.1–99.5) and 98.4% (990/1006, 95%CI 97.4–99.0) agreement with the reference method. Lastly, comparison in-between the predicted genetic AST results showed a high overall agreement of 99.2% (998/1006, 95% CI 98.4–99.6). When looking closer into the nine different antibiotics selected for this study, there were 100% agreement across the phenotypic AST and the predicted genotypic AST for the following antibiotics reported as susceptible (S) or resistant (R): clindamycin S, erythromycin S, trimethoprim S and R, tetracycline S and R, rifampicin S and R, vancomycin S and R and isoxazolyl penicillin R (Table 5). In total, there were 26 discrepancies across the bioinformatic workflows including 23 VME and three ME, where fusidic acid showed the greatest discordance with 11 VME followed by clindamycin and ciprofloxacin showing six and four VME respectively, as well as one ME for ciprofloxacin (Table 5). Furthermore, discordance across the bioinformatic workflows resulting in VME could also be observed for erythromycin (n = 2) and ME for isoxazolyl penicillin (n = 2) (Table 5). Looking closer into the 26 discrepancies from the bioinformatic workflows (Table 6) one can notice that the output i.e., identified genetic resistance markers or genes from the tools included in the 1928 and INH workflows sometimes differs.
Virulence genes and sequence type
We also compared the genetic predictions from the 1928 platform to those obtained by the INH concerning ST and presence (P) or absence (A) of selected virulence genes. In all, twenty-five genes were included in the analysis (Table 1). 1928 and the INH using VirulenceFinder (Fig. 2) showed overall 99.4% (1267/1275, 95% CI 98.7–99.7) agreement in genetic prediction of the chosen genes among the S. aureus (Table 7). The 1928 platform predicted the presence of another eight virulence genes than the INH using VirulenceFinder did, i.e., etA (n = 2), etB (n = 2) and tsst1 (n = 4) (Table 7).
In total, for 236 of the 250 isolates the results of MLST were consistent for the INH using MLST 2.0 (Fig. 2) and 1928 (Table 8). For 19 of the 255 isolates, the ST could not be definitely determined with either the INH or 1928. For three isolates, the ST was determined with 1928 but not with MLST 2.0 (SA117: ST50, SA656: ST5, and SA1723: ST39), while for two isolates, the ST was determined with MLST 2.0 but not with 1928 (SA331 and SA332, both ST6674). However, it can be noted that both workflows identified the same allelic profile for these two isolates even though 1928 did not yield any ST. Since typing of the SCCmec element is of importance to follow a possible MRSA outbreak, the 1928 platform also delivers a result about what kind of SCCmec element the S. aureus isolate is encoding. The 1928 platform predicted 17 isolates to encode for a SCCmec type, whereof four known and thirteen unknown types (Table 9). Among the four known SCCmec types, three of these were indeed predicted for the three isolates being phenotypic identified as MRSA in the clinical lab (SA 606, SA 1857 and SA 1153), but the clinical isolate, SA 1197, predicted to carry SCCmec XI was not. The 1928 platform genetically predicted the presence of mecC, while INH applying ResFinder predicted mecA, in the clinical isolate, SA 1197.
In this study, the 1928 platform was benchmarked to reference methods in the clinical laboratory as well as to an in-house developed bioinformatic pipeline (Fig. 2). Among all bacterial isolates collected during the prospective observational study of community-onset severe sepsis and septic shock , S. aureus was one of the most common etiological agents among the patients suspected of having sepsis.
Currently the INH includes the tool JSpeciesWS (Fig. 2), calculating the pairwise ANI against reference genomes  for prediction of species identification. Recently, another study , using similar preprocessing tools for FASTQ PE files retrieved from Illumina sequencing, also showed good achievement in species identification using the JSpeciesWS tool. Nevertheless, during the development of the INH, different tools for prediction of species identification were assessed i.e., 16S rRNA based species identification of S. aureus using CGE SpeciesFinder , kmer-based species identification with kmer size 16 and prefix “ATG” with CGE KmerFinder [35,36,37]. Also, species discrimination application of dDDH based on the Type (Strain) Genome Server, TYGS  was assessed. CGE Kmerfinder and TYGS showed high agreement with the reference method 99.2% (262/264), while SpeciesFinder predicted only 76.5% (202/264) as S. aureus (Additional file 2). Similar challenges have been reported earlier . The 1928 platform analysis for species identification, applying assembly free kmer-based method, showed 99.2% (262/264) agreement to the reference method. It should however be emphasized that depth/coverage of 11-29X were allowed for the species identification of nine isolates. Including only the FASTQ files passing 1928 internal quality threshold, the 1928 platform would have shown 100% (255/255) agreement to the reference method. Among the jointly predicted discrepant results (2/264), one isolate (SA 310) was predicted to be S. epidermidis by both JSpeciesWS and 1928, while the second isolate (SA 1413) was predicted as non-staphylococcus spp. by 1928, whereas JSpeciesWS predicted S. argenteus. In 2015, two novel species of the genus Staphylococcus were identified by WGS using the Illumina HiSeq platform, where one was proposed as S. argenteus . The general clinical impact of S. argenteus is difficult to assess because of the limited number of studies and datasets, and divergent observations exist, but recent studies suggest that the frequency of healthcare-associated infections, morbidity and mortality are comparable to those of S. aureus [40, 41]. In addition, there have been multiple reports of bloodstream infections among which S. argenteus methicillin resistant isolates have been isolated [40, 42,43,44], altogether illustrating the importance of S. argenteus identification. To date, classical routine diagnostics do not distinguish this species from S. aureus . Though, since April 2018 the clinical microbiology laboratory Unilabs, Skövde, is using the updated Bruker microorganism database MBT Compass Library DB-7854 (Bruker Daltonics, Germany) including identification of S. argenteus. Shortly after all WGS data collected in our study had been analyzed, the very first report of S. argenteus in Sweden was published and the 1928 platform was updated accordingly .
Antibiotic susceptibility test
Combined predicted genotypic antibiotic susceptibility from both of the bioinformatic workflows showed 98.0% (989/1006, 95% CI 97.3–99.0) agreement to phenotypic AST (Table 5), which has also been shown in other studies for S. aureus [9, 13, 46, 47]. The phenotypic AST results reported as resistant were less concordant than the phenotypic AST results reported as susceptible (Table 5). Similar results have been reported, when studying the accuracy of three different bioinformatic systems Genefinder, Mykrobe and Typewriter in genetic prediction of AST from S. aureus WGS data .
These bioinformatic systems showed challenges in concordant genetic predicted AST with phenotypic AST for the antibiotics, ciprofloxacin and fusidic acid, which was also the case in our study (Table 5). Among the 26 discrepancies across the bioinformatic workflows fusidic acid showed the greatest discordance with 11 VME followed by clindamycin and ciprofloxacin showing six and four VME respectively, as well as one ME for ciprofloxacin (Table 5). The discrepancies reported for the fusidic acid comes from the 1928 platform predicting the presence of fusC in three isolates (SA 535, SA 846 and SA 1162), while the INH did not predict the presence of any genes, resulting in three VME for the INH. Also, four isolates reported to be phenotypic resistant to fusidic acid (SA 1333, SA 1349, SA 1640 and SA 1370) were genotypically predicted susceptible since no gene or resistance markers could be predicted by the bioinformatic workflows (Table 6), resulting in four VME for each bioinformatic workflow. The resistance mechanism by fusidic acid, inhibiting protein synthesis, has been shown to have multiple genetic causes, some of which have only recently been discovered , illustrating the need for recognition of novel variants in the systems database for in silico prediction of resistance and susceptibility. For ciprofloxacin four isolates reported to be phenotypic resistant (SA 907, SA 1525, SA 1852 and SA 1153) the INH did not predict the presence of any gene, while the 1928 platform predicted the isolates to be resistant by the presence of several different genotypic resistance markers (Table 6), resulting in four VME for INH. Also, isolate SA 1046, reported to be phenotypic susceptible to ciprofloxacin was predicted by 1928 to be resistant by the presence of different genotypic resistance markers, while the INH could not predict any genes (Table 6), resulting in one ME for the 1928 platform. Even though only 1.0% of cases were reported in this study as phenotypically resistant for clindamycin, the bioinformatic workflows predicted all of them to be genotypically susceptible, resulting in six VME (Table 5). Bioinformatic tools showing concordant predictions for clindamycin but disagreed with phenotypic AST for S. aureus have also been reported in other studies [9, 49]. Likewise, previous studies of clindamycin resistance have reported positive ermC PCR results from nondetectable resistance phenotypes, suggesting that plasmids conferring resistance to these antibiotics may be lost in subculture, and therefore not present in the WGS data [46, 50]. ME reported for clindamycin may be inducible clindamycin resistant not detected by current phenotypic methods, but present in the WGS data. Since there has recently been evidence for increased worldwide inducible clindamycin resistance [51, 52], the bioinformatic workflows should consider this antibiotic group and continue the development of algorithm taking these identified challenges into account. For erythromycin, three isolates (SA 61, SA 63 and SA 804) were phenotypic resistant and both bioinformatic workflows predicted the presence of ermC in two isolates (SA 61 and SA 63), but the output from the bioinformatic workflows did not tell if it was plasmid mediated. The third isolate, SA 804, no resistance markers or genes was predicted by the bioinformatic workflows, resulting in two VME (Table 6). Since the antibiotic group fusidic acid had most VME (n = 11), the highest VME rate was also identified for fusidic acid, where 1928 reported 1.9% (4/206) and INH 3.4% (7/206), followed by clindamycin, where both bioinformatic workflows reported a VME rate of 1.4% (3/212), and for ciprofloxacin the 1928 platform showed a VME rate of 1.4% (1/70) and the INH 5.7% (4/70) (Table 5). Other studies using bioinformatic workflows such as blastn and tblast  and Next Gen Diagnostic  showed similar VME rates of 1.4% and 1.2% for ciprofloxacin respectively, while higher VME rates for clindamycin has been reported when using the 1928 platform i.e., 8.8% and 5.9% when using Next Gen Diagnostics . Discordant AST genotypic predictions could be due to different algorithms being employed by the bioinformatic tools, demonstrating the need of international agreement on quality control. Only data sets passing agreed quality control metrics should be used in antimicrobial susceptibility predictions as resistance genes or mutations otherwise might be missed in sequences of poor quality . In this study 100% (264/264) and 96.6% (255/264) FASTQ PE-files passed internal quality control metrics used by the INH and the 1928, respectively. The discordant AST genetic predictions for ciprofloxacin and fusidic acid was probably due to differences in the resistance database for the two bioinformatic workflows (Table 6). Nevertheless, the other discordant results cannot be deducted if the discordant AST genetic predictions were attributed to differences in the resistance database or the combination of assembly + BLAST within ResFinder 2.0 versus the assembly-free kmer-based method of 1928. A recent systematic review, using the CARD database 3.0.3 and Resfinder 4.0 on data retrieved from only Gram-negative bacteria, suggested the complexity of connecting genotype to phenotype with factors not yet considered in the resistance databases, for example gene regulation etc. . Individually, both 1928 and INH using Resfinder demonstrated high agreement with the phenotypic AST (Table 5). However, among the discordant results for each bioinformatic workflow, the 1928 platform showed lower VME rate than the INH using Resfinder, 0.8% (8/1006) versus 1.5% (15/1006), while the ME rate was slightly higher for the 1928 platform compared to the INH 0.2% (2/1006) versus 0.1% (1/1006). This is of importance, since VME, false negatives, might result in use of an ineffective therapeutic agent for treatment, leading to treatment failure, while a ME might limit therapeutic options and complicate treatment [14, 15].
Virulence genes and sequence type
Identification of S. aureus virulence genes can give the clinician insight into an infection’s pathogenesis and supporting the choice of therapy [54, 55]. The list of virulence genes that 1928 detects (Table 1) has been formed by requests from the platform's users. The exfoliative toxins, encoded by etA and etB are the cause for staphylococcal scalded skin syndrome . The tsst1 gene, encoding the toxic shock syndrome toxin-1, may cause staphylococcal toxic shock syndrome . The Panton-Valentine leucocidin (PVL) exotoxin, encoded by the lukF-PVL and lukS-PVL genes, is associated with S. aureus infections and is linked to infection severity and outcome in invasive disease . During the time the study took place the clinical lab did not perform any reference method for identification of these genes or expected phenotype, such as agglutination or ELISA assays for detection of toxic shock syndrome toxin-1 [59, 60]. Therefore, no benchmarking with results from the clinical lab can be addressed. The bioinformatic workflows showed overall high agreement in the genetic prediction of the virulence traits (Table 7). The S. aureus isolates collected during this study showed highest genetic predicted frequencies of the tsst1 and lukF-PVL, lukS-PVL genes (Table 7). The 1928 platform predicted the tsst1 and lukF-PVL, lukS-PVL to be present among 15.7% (40/255) and 2.7% (7/255) of the S. aureus isolates, while INH using VirulenceFinder predicted the tsst1 and lukF-PVL, lukS-PVL to be present among 14.1% (36/255) and 2.7% (7/255) of the S. aureus isolates. Another epidemiological marker is typing of the SCCmec element, aiding in understanding the evolution of MRSA and to follow a possible MRSA outbreak. The 1928 platform deliver a result about what kind of SCCmec element the S. aureus isolate is encoding, by including genetic prediction of thirteen different genes from the SCCmec casettes (Table 1). The platform predicted four known SCCmec types whereof the S. aureus isolate (SA 1197) predicted to belong to SCCmec type XI, have a cassette containing the recently identified mecC gene [61, 62] (Table 9). This isolate was identified as phenotypic susceptible to isoxazolyl penicillin, explaining the discordant result between the predicted genotypic AST results by both bioinformatic workflows (1/244, 0.4%) compared to the phenotypic AST (Table 5). In Sweden, the first MRSA with mecC was isolated in 2003 from a hedgehog but was not described as mecC until 2012 . Since 2012 the Swedish Communicable Diseases Act has been including S. aureus with mecC as a mandatory notifiable disease and handled in the same way as S. aureus with mecA regarding follow-up and contact tracing among household and healthcare contacts. Therefore, the current recommended routine diagnostics is to include a PCR assay for simultaneous detection of mecA and mecC  if a S. aureus isolate is resistant or intermediate resistant to the β-lactam cefoxitin. Nowadays, the prevalence of human mecC-MRSA infections is very low. However, mecC-MRSA isolate transmission between different hosts indicates the great capacity of these isolates for spreading and still the possible impact that these isolates can have in clinical settings remains unknown . In the SCCmec types annotated as unknown by 1928, the ccrA and ccrB genes were found, but mecA and mecC were absent. The absence of mecA/mecC agreed with the phenotypic AST which showed susceptibility to isoxazolyl penicillin of these isolates. It can be likely that these isolates harbor SCCmec remnants where they have lost the mecA and mecC genes, as has been observed in an earlier study . Another method being used to investigate the relationship between pathogens, but more on a global level, is by MLST and determination of ST. The 1928 platform showed overall 97.9% (231/236, 95% CI 95.0–99.2%) agreement with the INH using MLST 2.0 in predicting STs of the S. aureus isolates (Table 7). However, it should also be noted that for 19 isolates, the ST could not be definitely determined with the INH using MLST 2.0 or 1928. Comparison of classical MLST software for NGS data, have shown that not all MLST applications function as expected. MLST 2.0 was one of the tools used . Problems with some software included: poorly updated databases, computationally inefficient methods, false-positive results, inability to call alleles at low coverage and variable performance in the presence of mixed samples . Therefore, there is scope for improvement.
Time and user-friendliness
For WGS to be adopted in infection control and public health, it is required to be fast and generate robust results regarding the genomic context. Indeed, both of the bioinformatic workflows showed reliable results by demonstrating high agreement with the results retrieved in clinical routine, but there were differences in processing times between the bioinformatic workflows. The INH is code-level workflow, requiring formal bioinformatic support for operation and included steps of quality control followed by downstream analysis of the sequencing data. Estimated computational time required for analysis of one bacterial isolate, including two FASTQ PE files as input to the INH (Fig. 2) was 5–6 h using the Intel(R) Core (TM) i5-6300U CPU @ 2.40 GHz 2.40 GHz, RAM 16 GB, 64-bit PC. In more detail, the preprocessing of the FASTQ PE files took about 15 min, assembly and scaffolding about 2–3 h and finally annotation about 3 h, but time increased with queue size for the CGE webserver. The user needed to make a manual summary of the retrieved output. Using the same computer power, the computational time for the 1928 platform was 15–30 min. The raw sequencing data were directly uploaded and processed by the 1928 platform and the user received a summary of the retrieved output. Nevertheless, the FASTQ files that did not pass the internal quality control took about 24 h before the failed result was reported. Although being very user-friendly, a limitation with 1928 is that the user is restricted to the analyses included in the platform as opposed to the INH pipeline which can be extended with additional analyses available on the CGE and JSpeciesWS. There are also possibilities to extend the genotypic AST tools for the INH, since there are several freely accessible bioinformatics resources for detection of antimicrobial resistance determinants in DNA or amino acid sequence data, so far, e.g. ARG-ANNOT, CARD, SRST2, MEGARes, Genefinder, ARIBA, KmerResistance and AMRFinder .
Altogether, the benchmarking revealed that both bioinformatic workflows deliver results with high accuracy aiding diagnostics of severe infections caused by S. aureus, while the ST of S. aureus show scope for improvement. Our study is validating the performance of the 1928 platform in clinical routine for in silico species identification, antibiotic susceptibility testing and virulence profiling. The 1928 platform is also suitable for use in a clinical laboratory, since it is more user-friendly and deliver results timely. Still genotypic predictions cannot yet replace the phenotypic tests as in silico AST prediction for other organisms has been proved more challenging, especially for Gram-negative bacteria [19, 69] where the present understanding of genetic basis of resistance is less comprehensive. Additionally, the standardization of WGS workflows is a central requirement when entering the clinical diagnostics.
Availability of data and materials
The datasets generated and/or analysed during the current study are available in the online NCBI repository, https://www.ncbi.nlm.nih.gov/, BioProject PRJNA606666, http://www.ncbi.nlm.nih.gov/bioproject/606666
Antibiotic susceptibility testing
In-house developed bioinformatic pipeline
Multi-locus sequence typing
Very major error
Methicillin resistant S. aureus
European Committee on Antimicrobial Susceptibility Testing
Center for Genomic Epidemiology
Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet. 2012;380(9859):2095–128.
Bauer M, Reinhart K. Molecular diagnostics of sepsis–where are we today? Int J Med Microbiol. 2010;300(6):411–3.
Kumar A, Ellis P, Arabi Y, Roberts D, Light B, Parrillo JE, et al. Initiation of inappropriate antimicrobial therapy results in a fivefold reduction of survival in human septic shock. Chest. 2009;136(5):1237–48.
Garnacho-Montero J, Huici-Moreno MJ, Gutierrez-Pizarraya A, Lopez I, Marquez-Vacaro JA, Macher H, et al. Prognostic and diagnostic value of eosinopenia, C-reactive protein, procalcitonin, and circulating cell-free DNA in critically ill patients admitted with suspicion of sepsis. Crit Care. 2014;18(3):R116.
Valles J, Rello J, Ochagavia A, Garnacho J, Alcala MA. Community-acquired bloodstream infection in critically ill adult patients: impact of shock and inappropriate antibiotic therapy on survival. Chest. 2003;123(5):1615–24.
Fournier PE, Dubourg G, Raoult D. Clinical detection and characterization of bacterial pathogens in the genomics era. Genome Med. 2014;6(11):114.
Endrullat C, Glokler J, Franke P, Frohme M. Standardization and quality management in next-generation sequencing. Appl Transl Genom. 2016;10:2–9.
Ljungstrom L, Andersson R, Jacobsson G. Incidences of community onset severe sepsis, Sepsis-3 sepsis, and bacteremia in Sweden - a prospective population-based study. PLoS ONE. 2019;14(12): e0225700.
Mason A, Foster D, Bradley P, Golubchik T, Doumith M, Gordon NC, et al. Accuracy of different bioinformatics methods in detecting antibiotic resistance and virulence factors from staphylococcus aureus whole-genome sequences. J Clin Microbiol. 2018;56:9.
Petit RA 3rd, Read TD. Staphylococcus aureus viewed from the perspective of 40,000+ genomes. PeerJ. 2018;6: e5261.
Quijada NM, Hernandez M, Oniciuc EA, Eiros JM, Fernandez-Natal I, Wagner M, et al. Oxacillin-susceptible mecA-positive Staphylococcus aureus associated with processed food in Europe. Food Microbiol. 2019;82:107–10.
Kumar N, Raven KE, Blane B, Leek D, Brown NM, Bragin E, et al. Evaluation of a fully automated bioinformatics tool to predict antibiotic resistance from MRSA genomes. J Antimicrob Chemoth. 2020;75(5):1117–22.
Gordon NC, Price JR, Cole K, Everitt R, Morgan M, Finney J, et al. Prediction of Staphylococcus aureus antimicrobial resistance by whole-genome sequencing. J Clin Microbiol. 2014;52(4):1182–91.
Su M, Satola SW, Read TD. Genome-based prediction of bacterial antibiotic resistance. J Clin Microbiol. 2019;57:3.
Mahfouz N, Ferreira I, Beisken S, von Haeseler A, Posch AE. Large-scale assessment of antimicrobial resistance marker databases for genetic phenotype prediction: a systematic review. J Antimicrob Chemother. 2020;75(11):3099–108.
Tang Hallback E, Karami N, Adlerberth I, Cardew S, Ohlen M, Engstrom Jakobsson H, et al. Methicillin-resistant Staphylococcus argenteus misidentified as methicillin-resistant Staphylococcus aureus emerging in western Sweden. J Med Microbiol. 2018;67(7):968–71.
Giske CG, Dyrkell F, Arnellos D, Vestberg N, Panna SH, Froding I, et al. Transmission events and antimicrobial susceptibilities of methicillin-resistant Staphylococcus argenteus in Stockholm. Clin Microbiol Infec. 2019;25:10.
Enstrom J, Froding I, Giske CG, Ininbergs K, Bai XN, Sandh G, et al. USA300 methicillin-resistant Staphylococcus aureus in Stockholm, Sweden, from 2008 to 2016. PLoS ONE. 2018;13(11):67.
Saxenborn P, Baxter J, Tilevik A, Fagerlind M, Dyrkell F, Pernestig AK, et al. Genotypic characterization of clinical Klebsiella spp. isolates collected from patients with suspected community-onset sepsis. Sweden Front Microbiol. 2021;12:640408.
Ljungstrom L, Enroth H, Claesson BE, Ovemyr I, Karlsson J, Froberg B, et al. Clinical evaluation of commercial nucleic acid amplification tests in patients with suspected sepsis. BMC Infect Dis. 2015;15:199.
Enroth H, Retz K, Andersson S, Andersson C, Svensson K, Ljungstrom L, et al. Evaluation of QuickFISH and maldi Sepsityper for identification of bacteria in bloodstream infection. Infect Dis (Lond). 2019;789:1–10.
Andrews S. FastQC: a quality control tool for high throughput sequence data. https://www.bioinformatics.babraham.ac.uk/projects/fastqc/. 2010.
Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30(15):2114–20.
Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J Comput Biol. 2012;19(5):455–77.
Zankari E, Hasman H, Cosentino S, Vestergaard M, Rasmussen S, Lund O, et al. Identification of acquired antimicrobial resistance genes. J Antimicrob Chemother. 2012;67(11):2640–4.
Joensen KG, Scheutz F, Lund O, Hasman H, Kaas RS, Nielsen EM, et al. Real-time whole-genome sequencing for routine typing, surveillance, and outbreak detection of verotoxigenic Escherichia coli. J Clin Microbiol. 2014;52(5):1501–10.
Larsen MV, Cosentino S, Rasmussen S, Friis C, Hasman H, Marvig RL, et al. Multilocus sequence typing of total-genome-sequenced bacteria. J Clin Microbiol. 2012;50(4):1355–61.
Richter M, Rossello-Mora R. Shifting the genomic gold standard for the prokaryotic species definition. Proc Natl Acad Sci U S A. 2009;106(45):19126–31.
Ciufo S, Kannan S, Sharma S, Badretdin A, Clark K, Turner S, et al. Using average nucleotide identity to improve taxonomic assignments in prokaryotic genomes at the NCBI. Int J Syst Evol Microbiol. 2018;68(7):2386–92.
Enright MC, Spratt BG. Multilocus sequence typing. Trends Microbiol. 1999;7(12):482–7.
R: language and environment for statistical computing. [Internet]. R foundation for statistical Computing. 2020. http://www.R.project.org/.
Agresti A, Coull BA. Approximate is better than “exact” for interval estimation of binomial proportions. Am Stat. 1998;52(2):119–26.
Kluyver T, Ragan-Kelley B, Perez F, Granger B, Bussonnier M, Frederic J, et al. Jupyter Notebooks-a publishing format for reproducible computational workflows. In:Positioning and Power in Academic Publishing: Players, Agents and Agendas. 2016. p. 87–90.
Anaconda Documentation [Internet]. Anaconda documentation 2016. https://docs.anaconda.com/.
Larsen MV, Cosentino S, Lukjancenko O, Saputra D, Rasmussen S, Hasman H, et al. Benchmarking of methods for genomic taxonomy. J Clin Microbiol. 2014;52(5):1529–39.
Clausen ML, Agner T, Lilje B, Edslev SM, Johannesen TB, Andersen PS. Association of disease severity with skin microbiome and filaggrin gene mutations in adult atopic dermatitis. JAMA Dermatol. 2018;154(3):293–300.
Hasman H, Saputra D, Sicheritz-Ponten T, Lund O, Svendsen CA, Frimodt-Moller N, et al. Rapid whole-genome sequencing for detection and characterization of microorganisms directly from clinical samples. J Clin Microbiol. 2014;52(1):139–46.
Meier-Kolthoff JP, Goker M. TYGS is an automated high-throughput platform for state-of-the-art genome-based taxonomy. Nat Commun. 2019;10(1):2182.
Tong SYC, Schaumburg F, Ellington MJ, Corander J, Pichon B, Leendertz F, et al. Novel staphylococcal species that form part of a Staphylococcus aureus-related complex: the non-pigmented Staphylococcus argenteus sp. Nov. and the non-human primate-associated Staphylococcus schweitzeri sp. nov. Int J Syst Evol Microbiol. 2015;65(Pt 1):15–22.
Thaipadungpanit J, Amornchai P, Nickerson EK, Wongsuvan G, Wuthiekanun V, Limmathurotsakul D, et al. Clinical and molecular epidemiology of Staphylococcus argenteus infections in Thailand. J Clin Microbiol. 2015;53(3):1005–8.
Chen SY, Lee H, Wang XM, Lee TF, Liao CH, Teng LJ, et al. High mortality impact of Staphylococcus argenteus on patients with community-onset staphylococcal bacteraemia. Int J Antimicrob Agents. 2018;52(6):747–53.
Holt DC, Holden MT, Tong SY, Castillo-Ramirez S, Clarke L, Quail MA, et al. A very early-branching Staphylococcus aureus lineage lacking the carotenoid pigment staphyloxanthin. Genome Biol Evol. 2011;3:881–95.
Dupieux C, Blonde R, Bouchiat C, Meugnier H, Bes M, Laurent S, et al. Community-acquired infections due to Staphylococcus argenteus lineage isolates harbouring the Panton-Valentine leucocidin, France, 2014. Euro Surveill. 2015;20:23.
Argudin MA, Dodemont M, Vandendriessche S, Rottiers S, Tribes C, Roisin S, et al. Low occurrence of the new species Staphylococcus argenteus in a Staphylococcus aureus collection of human isolates from Belgium. Eur J Clin Microbiol. 2016;35(6):1017–22.
Becker K, Schaumburg F, Kearns A, Larsen AR, Lindsay JA, Skov RL, et al. Implications of identifying the recently defined members of the Staphylococcus aureus complex S argenteus and S schweitzeri: a position paper of members of the ESCMID Study Group for Staphylococci and Staphylococcal Diseases (ESGS). Clin Microbiol Infect. 2019;25(9):1064–70.
Holden MT, Hsu LY, Kurt K, Weinert LA, Mather AE, Harris SR, et al. A genomic portrait of the emergence, evolution, and global spread of a methicillin-resistant Staphylococcus aureus pandemic. Genome Res. 2013;23(4):653–64.
Koser CU, Holden MT, Ellington MJ, Cartwright EJ, Brown NM, Ogilvy-Stuart AL, et al. Rapid whole-genome sequencing for investigation of a neonatal MRSA outbreak. N Engl J Med. 2012;366(24):2267–75.
Hajikhani B, Goudarzi M, Kakavandi S, Amini S, Zamani S, van Belkum A, et al. The global prevalence of fusidic acid resistance in clinical isolates of Staphylococcus aureus: a systematic review and meta-analysis. Antimicrob Resist Infect Control. 2021;10(1):75.
Cunningham SA, Jeraldo PR, Schuetz AN, Heitman AA, Patel R. Staphylococcus aureus whole genome sequence-based susceptibility and resistance prediction using a clinically amenable workflow. Diagn Microbiol Infect Dis. 2020;97:3.
Bradley P, Gordon NC, Walker TM, Dunn L, Heys S, Huang B, et al. Rapid antibiotic-resistance predictions from genome sequence data for Staphylococcus aureus and Mycobacterium tuberculosis. Nat Commun. 2015;6:10063.
Abimanyu N, Murugesan S, Krishnan P. Emergence of methicillin-resistant Staphylococcus aureus ST239 with high-level mupirocin and inducible clindamycin resistance in a tertiary care center in Chennai, South India. J Clin Microbiol. 2012;50(10):3412–3.
Wang L, Liu Y, Yang Y, Huang G, Wang C, Deng L, et al. Multidrug-resistant clones of community-associated meticillin-resistant Staphylococcus aureus isolated from Chinese children and the resistance genes to clindamycin and mupirocin. J Med Microbiol. 2012;61(Pt 9):1240–7.
Ellington MJ, Ekelund O, Aarestrup FM, Canton R, Doumith M, Giske C, et al. The role of whole genome sequencing in antimicrobial susceptibility testing of bacteria: report from the EUCAST Subcommittee. Clin Microbiol Infect. 2017;23(1):2–22.
Kurlenda J, Grinholc M. Alternative therapies in Staphylococcus aureus diseases. Acta Biochim Pol. 2012;59(2):171–84.
Kane TL, Carothers KE, Lee SW. Virulence factor targeting of the bacterial pathogen Staphylococcus aureus for Vaccine and Therapeutics. Curr Drug Targets. 2018;19(2):111–27.
Bukowski M, Wladyka B, Dubin G. Exfoliative toxins of Staphylococcus aureus. Toxins. 2010;2(5):1148–65.
Spaulding AR, Salgado-Pabon W, Kohler PL, Horswill AR, Leung DY, Schlievert PM. Staphylococcal and streptococcal superantigen exotoxins. Clin Microbiol Rev. 2013;26(3):422–47.
Shallcross LJ, Fragaszy E, Johnson AM, Hayward AC. The role of the Panton-Valentine leucocidin toxin in staphylococcal disease: a systematic review and meta-analysis. Lancet Infect Dis. 2013;13(1):43–54.
Igarashi H, Fujikawa H, Shingaki M, Bergdoll MS. Latex agglutination-test for staphylococcal toxic shock syndrome Toxin-1. J Clin Microbiol. 1986;23(3):509–12.
Miwa K, Fukuyama M, Kunitomo T, Igarashi H. Rapid assay for detection of toxic shock syndrome Toxin-1 from human Sera. J Clin Microbiol. 1994;32(2):539–42.
Garcia-Alvarez L, Holden MT, Lindsay H, Webb CR, Brown DF, Curran MD, et al. Meticillin-resistant Staphylococcus aureus with a novel mecA homologue in human and bovine populations in the UK and Denmark: a descriptive study. Lancet Infect Dis. 2011;11(8):595–603.
Shore AC, Deasy EC, Slickers P, Brennan G, O’Connell B, Monecke S, et al. Detection of staphylococcal cassette chromosome mec type XI carrying highly divergent mecA, mecI, mecR1, blaZ, and ccr genes in human clinical isolates of clonal complex 130 methicillin-resistant Staphylococcus aureus. Antimicrob Agents Chemother. 2011;55(8):3765–73.
Monecke S, Gavier-Widen D, Mattsson R, Rangstrup-Christensen L, Lazaris A, Coleman DC, et al. Detection of mecC-positive Staphylococcus aureus (CC130-MRSA-XI) in diseased European hedgehogs (Erinaceus europaeus) in Sweden. PLoS ONE. 2013;8(6): e66166.
Pichon B, Hill R, Laurent F, Larsen AR, Skov RL, Holmes M, et al. Development of a real-time quadruplex PCR assay for simultaneous detection of nuc, Panton-Valentine leucocidin (PVL), mecA and homologue mecALGA251. J Antimicrob Chemother. 2012;67(10):2338–41.
Lozano C, Fernandez-Fernandez R, Ruiz-Ripa L, Gomez P, Zarazaga M, Torres C. Human mecC-Carrying MRSA: Clinical Implications and Risk Factors. Microorganisms. 2020;8:10.
Shore AC, Rossney AS, O’Connell B, Herra CM, Sullivan DJ, Humphreys H, et al. Detection of Staphylococcal Cassette Chromosome mec-Associated DNA Segments in Multiresistant Methicillin-Susceptible Staphylococcus aureus (MSSA) and Identification of Staphylococcus epidermidis ccrAB4 in both Methicillin-Resistant S aureus and MSSA. Antimicrobial Agents Chemother. 2008;52(12):4407–19.
Page AJ, Alikhan NF, Carleton HA, Seemann T, Keane JA, Katz LS. Comparison of classical multi-locus sequence typing software for next-generation sequencing data. Microb Genom. 2017;3(8): e000124.
Hendriksen RS, Bortolaia V, Tate H, Tyson GH, Aarestrup FM, McDermott PF. Using Genomics to Track Global Antimicrobial Resistance. Front Public Health. 2019;7:89.
Reuter S, Ellington MJ, Cartwright EJ, Koser CU, Torok ME, Gouliouris T, et al. Rapid bacterial whole-genome sequencing to enhance diagnostic and public health microbiology. JAMA Intern Med. 2013;173(15):1397–404.
Open access funding provided by University of Skövde. Swedish Knowledge Foundation BioMine Grant No. 206/0330, Swedish Knowledge Foundation Associate Senior Lecturer in Systems biology, Grant No. 2017/14, Stiftelsen Tornspiran, Internal research fund, Unilabs AB.
Ethics approval and consent to participate
All methods were performed in accordance with the relevant guidelines and regulations. The study was approved by the Regional Ethical Review Board at the University of Gothenburg (permit 376–11). Furthermore, as the present study only focused on bacterial isolates recovered from cultures included in the routine patient care, individual informed consent is deemed unnecessary according to national regulations (2003:460).
Consent for publication
HE was employed by the company Unilabs. DA was employed by the company 1928 Diagnostics. DT, AP, AT, MS and SJ do not have any competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Additional file 1.
Specification of antibiotics used in the phenotypic antibiotic susceptibility testing for S. aureus isolates as part of the routine practice in the clinical laboratory. Phenotypic antibiotic susceptibility testing results reported in this study are limited from the set of antibiotics included in the 1928 platform.
Additional file 2.
Genetically predicted species identification using bioinformatic tools for the 264 isolates identified as S. aureus by MALDI-TOF MS.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Shemirani, M.I., Tilevik, D., Tilevik, A. et al. Benchmarking of two bioinformatic workflows for the analysis of whole-genome sequenced Staphylococcus aureus collected from patients with suspected sepsis. BMC Infect Dis 23, 39 (2023). https://doi.org/10.1186/s12879-022-07977-0
- Whole-genome sequencing
- Antimicrobial susceptibility
- S. aureus
- Species identification
- Virulence genes
- Clinical microbiology
- Illumina sequencing