Featured Publication
Staphylococcus epidermidis joint isolates: Whole-genome sequencing demonstrates evidence of hospital transmission and common antimicrobial resistance
We investigated genetic, epidemiologic, and environmental factors contributing to positive Staphylococcus epidermidis joint cultures…In total, 60 phenotypically distinct S. epidermidis isolates were identified. After removal of duplicates and impure samples, 48 isolates were used for the phylogenomic analysis, and 45 (93.7%) isolates were included in the clonality analysis.
View PublicationPublications
Staphylococcus epidermidis joint isolates: Whole-genome sequencing demonstrates evidence of hospital transmission and common antimicrobial resistance
We investigated genetic, epidemiologic, and environmental factors contributing to positive Staphylococcus epidermidis joint cultures…In total, 60 phenotypically distinct S. epidermidis isolates were identified. After removal of duplicates and impure samples, 48 isolates were used for the phylogenomic analysis, and 45 (93.7%) isolates were included in the clonality analysis.
View PublicationRalstonia pickettii and Pseudomonas aeruginosa Bloodstream Infections Associated with Contaminated Extracorporeal Membrane Oxygenation Water Heater Devices
After tracking a rash of unusual infections, Brigham and Women’s Hospital’s infection control team called on epiXact, our rapid, whole-genome sequencing (WGS) service using single-nucleotide polymorphism analysis, to help trace outbreak strains of Burholderia, Ralstonia picketti and Pseudomonas aeruginosa. With the high resolution offered by WGS, the infection control team was able to conclusively link the outbreak with (ECMO) water heater devices which would have been otherwise challenging due to the prevalence of multiple species collected from various locations.
View PublicationCluster of Burkholderia cepacia Complex Infections Associated with Extracorporeal Membrane Oxygenation Water Heater Devices
Our epiXact healthcare-associated infection (HAI) service helped Brigham and Women’s hospital link a cluster of Burkholderia cepacia complex infections in cardiothoracic ICU patients to contaminated Extracorporeal Membrane Oxygenation Water Heater Devices.
View Publication
Shiga Toxin–Producing Escherichia coli Transmission
via Fecal Microbiota Transplant
Our epiXact service for healthcare-associated infections (HAIs) confirmed the first known report of an undetected transmission of Shiga toxin-producing E. coli (STEC) in a fecal microbiota transplantation, despite enzyme-based STEC screening having been performed on donor samples. Following epiXact’s actionable findings, OpenBiome worked, in consultation with the FDA, to implement prospective PCR-based testing to enhance patient safety and avoid future transmissions.
View PublicationCommunity-acquired in name only: A cluster of carbapenem-resistant Acinetobacter baumannii in a burn intensive care unit and beyond
Mass General Hospital used our epiXact service to rapidly identify and respond to a highly-resistant A. baumannii outbreak in an ICU burn unit that was initially believed to be caused by community-transmission.
View PublicationPlasmids and genes contributing to high-level quinolone resistance in Escherichia coli
This research on an E. coli strain with a remarkably high resistance to broad-spectrum antibiotics helps us better understand the transmission of antibiotic resistance between strains via plasmids.
View PublicationDrug-Resistant E. coli Bacteremia Transmitted by Fecal Microbiota Transplant
Our epiXact service was used by Mass General Hospital to provide high-resolution whole genome sequencing analysis in less than two days to help uncover the cause of the first known fecal matter transplant patient death.
View PublicationMultiple Copies of qnrA1 on an IncA/C2 Plasmid Explain Enhanced Quinolone Resistance in an Escherichia coli Mutant
View Publication
Abstracts and Posters
Rapid species detection and antibiotic susceptibility profiling for bloodstream infections directly from whole blood at single digit pathogen concentrations
We successfully identified both bacterial and fungal BSI pathogens at single digit concentrations with high accuracy. The ability of Blood2BacTM to rapidly recover whole pathogen genomes and for Keynome® to provide accurate ID and AST predictions demonstrates their potential to have a high impact on the clinical management of BSIs utilizing whole genome sequencing.
Download PDFLarge-Scale Evaluation of AST Prediction using Resistance Marker Presence/Absence vs. Machine Learning on WGS Data
We assessed ResFinder, a publicly available bioinformatics tool that detects a curated set of resistance genes and point mutations, to test the utility of resistance marker presence or absence for predicting resistance/susceptibility on 107 species/drug combinations relevant for bloodstream infections. We also tested the performance of Keynome gAST, a machine learning method we developed that predicts AST from WGS data using the entirety of the bacterial genome without limiting to known resistance markers. Performance was assessed on >36,000 bacterial strains from MicrohmDB®, a large-scale database containing WGS data and phenotypic AST results for tens of thousands of clinical bacterial isolates.
Download PDFKeynome Diagnostic Pipeline
We demonstrate accuracy of Keynome gAST across a dataset of isolate strain genomic sequencing data, showing high accuracy across 68 species-drug combinations. Additionally, we show Keynome gAST accuracy on 100 contrived direct-from-blood (bacterial spike-in) samples, achieving >90% agreement with phenotypic AST across a panel of >40 pathogen-drug combinations. Finally, we demonstrate high accuracy on a limited dataset demonstrating Keynome gAST accuracy on polymicrobial contrived samples with 2 bacterial strains spiked into whole blood at low concentrations.
Download PDFDevelopment of a Culture-Free Diagnostic for Urosepsis Leveraging Whole Genome Sequencing and Machine Learning
Complicated urinary tract infections (cUTIs) can be life-threatening and occur most often in hospitalized patients. There are 2.8 million cases of cUTIs in the U.S. each year, with over 20/% of cases progressing to urosepsis, causing nearly 150,000 deaths annually. Effective clinical management of cUTIs requires rapid identification (ID) of causative pathogens and reliable antibiotic susceptibility tests. Urine culture, the current gold-standard, exhausts vital turn-around time from specimen collection to actionable information for treatment. Day Zero Diagnostics (DZD) developed DZD-UroSeq to address the need for a culture-free diagnostic, leveraging ultra-high enrichment of pathogen DNA, whole-genome sequencing (WGS), and machine learning to deliver high resolution species ID with antimicrobial resistance and susceptibility profiling from patient urine in under 6 hours.
Download PDFIdentification and Antibiotic Resistance Profiling of Bacterial Pathogens Directly from Whole Human Blood Utilizing Blood2Bac
Bloodstream infections (BSI) are a leading cause of patient mortality with early diagnosis and pathogen identification key to improving clinical outcomes. However, longer turn-around times associated with culture-based diagnostics limit the ability of clinical labs to provide timely information to clinicians. Moreover, delayed treatment for patients significantly impacts clinical outcomes with septic shock mortality rates increasing by 8% per hour without appropriate antibiotics. Day Zero Diagnostics has developed Blood2Bac™, a culture-free, pathogen-agnostic technology which enriches bacterial pathogens directly from whole human blood and utilizes bacterial whole genome sequencing and proprietary algorithms to provide sensitive identification of pathogens and antimicrobial resistance (AMR) information at single digit CFU/mL levels.
Download PDFDirect-from-Blood Microbial Sequencing Assay for Pathogen and Antibiotic Resistance Detection in Bloodstream Infections
We developed Blood2BacTM, a novel method for species agnostic ultrahigh enrichment (UHE) of bacteria directly from whole blood that depletes human DNA by 8-10 orders of magnitude. Blood2Bac is followed by whole genome sequencing (WGS) on an Oxford Nanopore Technologies sequencing device for rapid data generation. WGS data is analyzed with Keynome® ID, our algorithm for species pathogen identification in mixed human/pathogen samples, and Keynome® gAST (genomic antibiotic susceptibility testing), our machine learning algorithm for determining AST from genomic sequences to accurately predict resistance. With an end-to-end turnaround time of 8 hours, we demonstrate comprehensive capabilities for whole genome recovery, species identification, and AST determination direct from blood at low input concentrations.
Download PDFKeynome gAST: a machine learning system for predicting antimicrobial resistance phenotypes from whole-genome sequencing
Here we review the current state of diagnostics for bacterial bloodstream infections and present Keynome gAST (genomic AST) – a machine learning (ML) system for predicting AMR/S phenotype from pathogen whole-genome sequencing data derived directly from whole blood via an ultra-high pathogen DNA enrichment process. The ML models at the heart of the system are trained on MicrohmDB – our extensive database of paired pathogen genomes and traditional AST results. The models learn genomic signatures that predict AMR/S phenotype, even for species and drugs where resistance mechanisms are not yet well characterized. We show Keynome gAST can achieve >90% agreement with traditional AST (>95% for strongly susceptible/resistant samples) on 100 contrived direct-from-blood samples across a panel of >40 pathogen-drug combinations, and demonstrate its superiority to resistance marker based approaches.
Download PDFAccurate species identification and antibiotic susceptibility prediction for multiple pathogens in contrived polymicrobial blood samples using whole genome Oxford Nanopore sequencing
Day Zero Diagnostics® is developing a diagnostic for identifying bacteria in bloodstream infections using ultra-high pathogen enrichment (Blood2Bac™) followed by whole genome sequencing of bacteria on the Oxford Nanopore Technologies (ONT) platform. Together with Keynome® ID, our species identification algorithm, our end-to-end process can detect the correct species in whole blood in hours, compared to blood culture, which can often take days. Keynome® g-AST (genomic Antibiotic Susceptibility Testing), our proprietary machine learning algorithm, predicts antibiotic resistance using whole genome sequencing data.
Download PDFepiXact: Robust bacterial relatedness and outbreak detection pipeline for WGS data
DZD’s Illumina-based commercial HAI sequencing and analysis service, is used by multiple partnering hospitals to investigate a wide variety of suspected outbreaks. When infection control (IC) suspects transmission, cultured bacterial samples are sent to DZD. The epiXact test is a “rule-in/rule-out” test of an infection transmission event based on clonality of 2 or more bacterial isolates. This test is designed as a stand-alone test to provide clinicians with a definitive measure of the relatedness of bacterial isolates. Results are reported back in 1-2 days allowing IC to use the definitive genomic evidence to inform decisions regarding ward cleaning, staff screening, and equipment contamination.
Download PDFIdentification of subclinical healthcare-associated clusters of Staphylococcus epidermidis in an orthopedic patient population
Results presented by New England Baptist Hospital (NEBH) on the use of our epiXact PRO service to investigate genetic, epidemiologic, and environmental factors contributing to positive S. epidermidis joint cultures and prosthetic joint infection. S. epidermidis isolates from hip or knee cultures were identified and obtained between 2017-2020 from patients with one or more prior intraarticular procedures from NEBH. Whole-genome sequencing and single nucleotide polymorphism (SNP) based clonality analysis was performed using the epiXact service. This included species identification, in silico multi-locus sequence typing (MLST), phylogenomic analysis, along with genotypic assessment of the prevalence of specific antibiotic resistance and virulence genes.
Download PDFValidation of epiXact: Robust Bacterial Relatedness and Outbreak Detection Pipeline for WGS Data ECCMID 2022
Large-scale validation for Clinical Laboratory Improvement Amendments (CLIA ) certification for epiXact, our automated computational pipeline for detecting pathogen relatedness from WGS data. The epiXact pipeline demonstrated high accuracy for determining clonality between bacterial isolates across 5 species achieving 100% analytical sensitivity and 98.5% analytical specificity in determining clonality, and 100% repeatability.
Download PDFHigh concordance between short and long read sequencing for genomics-based species identification and antimicrobial resistance
This poster demonstrates the sustainability of ONT sequencing in the diagnosis and identification of bacterial infections and potential microbial resistance. We assessed the performance of Illumina short-read versus ONT long-read whole genome sequencing (WGS) data for species identification (ID) and antimicrobial susceptibility test (AST). The results were a high degree of concordance for ID (99.4%) and AST (97.7%) between the two sequencing platforms. Advances in nanopore sequencing have the potential to transform infectious disease diagnosis, cutting costs and leading to faster, more efficient treatments.
Download PDFSame-day transmission analysis of nosocomial transmission using nanopore whole-genome sequencing
This poster demonstrates the utility of the Oxford Nanopore Technologies (ONT) platform, a rapid sequencing technology, for use with our epiXact healthcare-associated infection (HAI) service. ONT sequencing offers many advantages with faster speed and lower costs over short-read technologies that could prove to be an attractive platform for the commercial epiXact service reducing turnaround time from ≥34-46 hours to same day service and pave the way for real-time HAI transmission detection and prospective outbreak warning system.
Download PDFUncluttering Case Clusters: Use of Rapid Sequencing to Exclude Transmission Events
Results presented by Mass General Hospital (MGH) on the use of epiXact healthcare-associated infection (HAI) service to investigate a cluster of methicillin-resistant Staphylococcus aureus (MRSA) and a cluster of carbapenem-resistant Enterobacterales (CRE) inpatient nosocomial infections. Whole-genome sequencing results allowed for early discontinuation of cluster investigations and conservation of resources.
Download PDFKeynome g-AST: Development of a Novel Machine Learning Method for Determining Bacterial Antibiotic Susceptibility from Genomic Sequences
This poster presents Keynome g-AST (genomic-AST), our data-driven approach to determining antibiotic susceptibility from genomic sequences. Keynome g-AST was trained and validated using our MicrohmDB dataset that consists of the whole genome sequences of over 45,000 microbial pathogens matched with their phenotypically derived AST results. Results show that Keynome g-AST is a promising method for determining the AST profile of pathogens and enable rapid culture-free AST diagnostics.
Download PDFepiXact : Rapid, precise and robust bacterial relatedness and outbreak detection from WGS data
This poster analyzes recent results of our epiXact service for healthcare-associated infections (HAIs). The poster highlights results from 24 recent cases of suspected HAIs submitted for epiXact investigation by clinical customers, encompassing a total of 116 bacterial samples across 12 different pathogens. The analysis demonstrated epiXact’s robust ability to detect outbreaks quickly, leveraging the automated and precise algorithm to provide conclusive evidence of 16 outbreak cases within 24-48 hours from sample receipt. With these rapid results, infection control specialists can make timely and accurate decisions to get the infection outbreak back under control.
Download PDFPilot study of a novel whole-genome sequencing based rapid bacterial identification assay in patients with bacteremia
This is the first proof-of-concept feasibility study in an inpatient clinical setting of our culture-free, species agnostic process using whole genome sequencing and algorithmic tools for identifying the species and antimicrobial resistance profile of a bloodstream infection in real-time. The results suggest the approach is potentially more sensitive and significantly faster than gold standard culture-based diagnostics.
Download PDFBlood2Bac: species ID and AMR prediction of bacterial pathogens at low concentrations in blood using a rapid ultra-high enrichment process and nanopore sequencing
This study demonstrated Blood2Bac’s ability, when paired with Keynome, to test for a broad range of infections (50 bacterial species) and recover almost the entirety of the pathogen’s genome directly from whole blood. The results represent a significant advancement over current molecular diagnostic approaches that are limited to testing just a handful of species and recover a very small portion of the pathogen’s genetic code.
Download PDFBacDetect: Development of a rapid, ultra-sensitive platform for bacterial detection from blood
To address slow turnaround times related to culture-based diagnostics for blood stream infections, we have developed BacDetect, a rapid DNA amplification-based detection method that determines the presence of gram-positive or gram-negative bacteria in a sample in 30 minutes.
Download PDFDemocratizing Sequencing for Infection Control: A Scalable, Automated Pipeline for WGS Analysis for Outbreak Detection
This abstract describes how our technology stack for detecting healthcare-associated infections (HAIs) can efficiently analyze large-scale datasets. Including a recent analysis of over 5,000 clinical bacterial samples collected between 2015 and 2019 which uncovered previously unidentified transmission clusters. With this robust technology stack, we can scale to analyze tens of thousands of samples and support infection control teams in performing prospective outbreak detection without the need for specialized computational biology training.
Download PDFRapid ultra-high enrichment of bacterial pathogens at low concentration from whole blood for species ID and AMR prediction using Oxford Nanopore sequencing
Blood2BacTM is our species-agnostic culture-free method for enriching bacterial DNA directly from whole blood samples by a factor of 100,000,000. When this is coupled with rapid whole genome sequencing and our algorithmic tools, we can determine bacterial species identification and antimicrobial resistance within hours instead of days with culture.
Download PDFMulti-copy qnrA1 Plasmid Causes Elevated Quinolone Resistance in E. coli
Analyzing the role large multi-drug resistance (MDR) plasmids play in cultivating resistance can be tremendously challenging. In this work, we leverage nanopore long-read sequencing to overcome these challenges and produced a complete sequence of an MDR plasmid with multiple copies of resistance genes that appeared to significantly increase resistance to ciprofloxacin in E. coli strains carrying the plasmid.
Download PDFAchievement of rapid whole genome coverage of bacterial pathogens at 1 CFU/mL in blood
Using whole genome sequencing for culture-free diagnosis of bacterial bloodstream infections is difficult because very little bacterial DNA is present in a clinical sample. Human DNA outnumbers bacterial DNA by 8-9 orders of magnitude. This work describes how we are able to rapidly enrich and sequence bacterial DNA for diagnosing bacterial bloodstream infections directly from clinical blood samples.
Download PDFcounterr: Characterization of Context Dependent MinION Sequencing Errors
In this work we described how we used counterr, our lightweight command line tool to characterize the error distributions in both amplified and native microbial ONT MinION sequencing data. Our results confirm a widely held belief that errors in MinION data strongly depend on sequence context. We hope that this improved error characterization can be useful for read error correction.
Download PDFQuantification of Predictive Power of Genomic Resistance Locus Databases Reveals Potential Limitations of Marker-Based Diagnostics
Tools that rapidly and accurately predict the antibiotic resistance of a clinical infection promise to allow healthcare providers to quickly provide targeted, effective treatments to patients. However, most emerging diagnostic approaches detect the presence or absence of only a limited panel of resistance markers. In the this work we analyze the power of genomic resistance locus databases to predict the resistance profiles of Staphylococcus aureus, Streptococcus pneumonia, and Mycobacterium tuberculosis clinical isolates.
Download PDF