# Prospective modeling and estimating the epidemiologically informative match rate within large foodborne pathogen genomic databases

## TL;DR

This paper studies how often genetic matches occur between patient and non-patient isolates in foodborne pathogen databases to improve public health surveillance.

## Contribution

The study introduces a model to estimate and predict match rates in genomic databases, emphasizing the need for non-clinical isolates.

## Key findings

- Match rates vary significantly across pathogens, with Salmonella having the highest at 46%.
- Logistic regression modeling shows good performance in predicting match rates based on database features.
- The study highlights the importance of including non-clinical isolates to improve match identification.

## Abstract

Much has been written about the utility of genomic databases to public health. Within food safety these databases contain data from two types of isolates—those from patients (i.e., clinical) and those from non-clinical sources (e.g., a food manufacturing environment). A genetic match between isolates from these sources represents a signal of interest. We investigate the match rate within three large genomic databases (Listeria monocytogenes, Escherichia coli, and Salmonella) and the smaller Cronobacter database; the databases are part of the Pathogen Detection project at NCBI (National Center for Biotechnology Information).

Currently, the match rate of clinical isolates to non-clinical isolates is 33% for L. monocytogenes, 46% for Salmonella, and 7% for E. coli. These match rates are associated with several database features including the diversity of the organism, the database size, and the proportion of non-clinical BioSamples. Modeling match rate via logistic regression showed relatively good performance. Our prediction model illustrates the importance of populating databases with non-clinical isolates to better identify a match for clinical samples. Such information should help public health officials prioritize surveillance strategies and show the critical need to populate fledgling databases (e.g., Cronobacter sakazakii).

The online version contains supplementary material available at 10.1186/s13104-024-06847-z.