DrPRG

DrPRG predicts drug resistance in Mycobacterium tuberculosis from whole-genome sequencing data by leveraging bacterial reference graphs to support genotypic drug susceptibility testing.

Key Features:

• Reference Graph Methodology: Uses the bacterial reference graph method Pandora to construct a Mycobacterium tuberculosis drug resistance reference graph from a global dataset of isolates, capturing common and rare haplotypes.
• Mutation Detection: Encodes genetic variation in the reference graph to detect known resistance-causing mutations and gene deletions.
• Benchmarking Performance: Demonstrated improved sensitivity and specificity for certain drugs in comparative studies of 44,709 Illumina and 138 Nanopore samples versus Mykrobe and TBProfiler, while using less memory and exhibiting faster runtimes except compared to Mykrobe on Nanopore data.
• Novel Insights: Identifies novel resistance-conferring genetic variations, including gene deletions such as katG and pncA.
• Potential for Reclassification: Suggests specific mutations that may warrant reclassification as resistance-associated variants.

Scientific Applications:

• Genotypic Drug Susceptibility Testing: Provides genotypic predictions of drug susceptibility to complement phenotypic DST using whole-genome sequencing data.
• Surveillance and Outbreak Management: Supports surveillance and management of tuberculosis outbreaks, including multidrug-resistant strains, by integrating genomic and phenotypic profiles.
• Research into Resistance Mechanisms: Enables investigation of the genetic basis of antimicrobial resistance by identifying novel variants and candidate reclassifications.

Methodology:

Constructs a Mycobacterium tuberculosis drug resistance reference graph using Pandora from a global isolate dataset, encodes variants and haplotypes in the graph, detects resistance mutations and gene deletions from Illumina and Nanopore whole-genome sequencing data, and was benchmarked against Mykrobe and TBProfiler across 44,709 Illumina and 138 Nanopore samples to assess sensitivity, specificity, runtime, and memory.