Pan-genomic and Polymorphic Driven Prediction of Antibiotic Resistance inElizabethkingia

Abstract

2The Elizabethkingia are a genetically diverse genus of emerging pathogens that exhibit multidrug 3 resistance to a range of common antibiotics. Two representative species, Elizabethkingia 4 bruuniana and Elizabethkingia meningoseptica, were phenotypically tested to determine 5 minimum inhibitory concentrations for five antibiotics. Ultra-long read sequencing with Oxford 6 Nanopore Technologies and subsequent de novo assembly produced complete, gapless circular 7 genomes for each strain. Alignment based annotation with Prokka identified 5,480 features in E. 8 bruuniana and 5,203 features in E. meningoseptica, where none of these identified genes or gene 9 combinations corresponded to observed phenotypic resistance values. Pan-genomic analysis, 10 performed with an additional 19 Elizabethkingia strains, identified a core-genome size of 11 2,658,537 bp, 32 uniquely identifiable intrinsic chromosomal antibiotic resistance core-genes 12 and 77 antibiotic resistance pan-genes. Using core-SNPs and pan-genes in combination with six 13 machine learning algorithms, binary classification of clindamycin and vancomycin resistance 14 achieved f1 scores of 0.94 and 0.84 respectively. Performance on the more challenging 15 multiclass problem for fusidic acid, rifampin and ciprofloxacin resulted in f1 scores of 0.70, 0.75 16 and 0.54 respectively. 17 18 19 101Mueller-Hinton Agar (MHA) and incubating for 24 hr (37°C). The MBC was determined as the 102 lowest antimicrobial concentration in which no visual colonies were observed. 103 Library preparation 104DNA libraries were prepared separately for each Elizabethkingia isolate following the 105 procedures outlined for the SQK-LSK208 2D sequencing kit (Oxford Nanopore Technologies, 106 United Kingdom) with the following protocol adjustments. A total of 1.5 µg of gDNA was 107 sheared in g-tubes (Covaris) at 4200 RPM for a targeted fragment size of 20 kb. End-repair was 108 6 performed following the manufacturer's recommended protocol for Ultra II End-prep enzyme 109 mix (NEB). Adapter ligation reaction incubations were increased to 15 minutes. All bead clean-110 ups used 0.4x AMPureXP beads (Beckman Coulter, Brea, CA) for additional size selection and 111 elutions were performed at 37°C for 20 minutes. DNA concentration of the library was 112 quantified using Quant-IT PicoGreen® dsDNA Assay Kit (ThermoFisher Scientific), measured 113 on Synergy H1, hybrid multi-mode microplate reader (BioTek). Final DNA library yields were 114 above the recommended 200 ng. 115 Single molecular real time sequencing 116 Two R9.4 flow cells were prepared for two corresponding MinIONs, each connected to a 117 separate Windows PC using a USB 3.0 connection. MinKNOW GUI application 1.0.8.0 from 118 Oxford Nanopore Technologies (ONT) was used to validate the MinION connection and to 119 monitor basic hardware details, like the number of active pores within each flow cell during 120 sequencing runs. Pore count validation was completed beforehand, with the Platform QC 121 command in MinKNOW. Flow cell prim...

Genomic epidemiology of Escherichia coli isolates from a tertiary referral center in Lilongwe, Malawi

“…Each assembly was aligned using minimap2 to databases of known serotypes, fimH type, genomic and plasmid sequence types including plasmid incompatibility, virulence factors, and antimicrobial resistance genes, listed in Table S1 (available in the online version of this article) [35][36][37][38][39][40]. Matches were made for each assembly using similar cutoffs to those used by the Centre for Genomic Epidemiology (CGE; https:// cge.…”

Genomic Epidemiology ofEscherichia coliIsolates from a Tertiary Referral Center in Lilongwe, Malawi

“…[31,32] Typing of isolates : Each assembly was aligned using minimap2 to databases of known serotypes, genomic and plasmid sequence types including plasmid incompatibility, virulence factors, and antimicrobial resistance genes, listed in Supplemental Table 1 . [33][34][35][36][37][38] Matches were made for each assembly using similar cutoffs to those used by the Center for Genomic Epidemiology (CGE; https://cge.cbs.dtu.dk/services/data.php ) tools, where at least 60% of the feature must match at >90% sequence identity.…”

Machine learning in the clinical microbiology laboratory: has the time come for routine practice?

“…Still, poor performances were observed for ceftazidime (81% for predicting resistance and 83% for predicting susceptibility) [53]. Similar approaches were described for Clostridioides difficile [54], E. coli [55] or Elizabethkingia species [56]. Besides genomic data, a number of ML systems analysed MALDI-TOF MS data to extend its potential for diagnostics, e.g.…”