Digital High Resolution Melt and Machine Learning for Rapid and Specific Diagnosis in Neonatal Sepsis

点击翻译按钮获取中文摘要

Project Summary Blood culture sensitivity in neonates is poor but is the “Gold Standard” for the diagnosis of sepsis. Universal genotyping of pathogen genomic sequences using High Resolution Melt (U-HRM) provides a simple, low cost, rapid, and modern alternative to blood culture testing. By measuring the fluorescence of an intercalating dye as PCR-amplified pathogen DNA fragments are heated and disassociate, sequence defined melt curves are generated with single-nucleotide resolution in a closed-tube reaction. We have advanced U- HRM into a digital PCR format (U-dHRM), where DNA sequences that are present in mixtures are individually amplified and identified as is needed for polymicrobial infections. We have also established unique signature melt curves for 37 bacterial species that commonly infect older children and adults and automatically identify them using machine learning technology. With the goal of creating an accurate and valid test for the timely diagnosis of neonatal sepsis, we will advance this technology to identify unique fungal, viral, and bacterial HRM signatures along with antibiotic resistance genes with an accuracy of 99-100% on minimal blood volume (1mL). Our aims are: Aim 1. Optimize and assess the U-dHRM platform for neonatal bacteremia diagnosis by expand our bacterial database (13 additional bacteria) to detect causes of >99% of neonatal bacterial infections, expand our antibiotic resistance gene database to include five clinically actionable genes, and assessing the performance of the system for bacteremia diagnosis in mock and clinical whole blood samples; Aim 2. Advance the U-dHRM platform for simultaneous detection of fungal and viral pathogens by upgrading our optical system to enable expansion to fungal and viral detection in a high-throughput format, multiplexing the assay to expand to viral and fungal pathogens causing >99% non-bacterial infections, and conducting analytical validation of the multiplexed platform using mock whole blood samples; and Aim 3. Advance the machine learning algorithm for detection of emerging pathogens by developing and integrating an anomaly detection algorithm for reporting emerging pathogens that are not included in our database and validating the algorithm using data generated in Aims 1 and 2. Thus, this proposal directly addresses the funding call by applying a multidisciplinary approach to overcome the biomedical challenge of rapidly diagnosis sepsis, a hidden public health disaster.