FACT: Salmonella Typing and Phenotypic Prediction From Genomes and Metagenomes Using Population Genomics and Machine Learning

This project aims to develop cyberinformatic tools to type Salmonella strains and predict resistance to sanitizers and antibiotics from genome and metagenome samples. This long-term goal will be achieved by the development of several intermediate objectives.Identify reference single nucleotide polymorphism (SNP) sites from ~180,000 genomes and develop a novel machine learning method to represent the SNP data which reduces the number of features and is tolerant to missing data. This tolerance is essential for identifying Salmonella in metagenomic samples.Develop deep learning models to accurately type metagenomic samples, predicting core gene Multi Locus Sequencing Type (cgMLST), whole genome Multi Locus Sequencing Type (wgMLST) and serotype.Develop deep learning models to predict the minimum inhibitory concentration (MIC) of antibiotics and sanitizers from both genomes and metagenomics samples.Create software to rapidly process reads or assembled genomes and metagenomes, typing the strains present and phenotypically characterizing them.