Biological machine learning combined with bacterial population genomics reveals common and rare allelic variants of genes to cause disease

Abstract

Highly dimensional data generated from bacterial whole genome sequencing is providing unprecedented scale of information that requires appropriate statistical frameworks of analysis to infer biological function from bacterial genomic populations. Application of genome wide association study (GWAS) methods is an emerging approach with bacterial population genomics that yields a list of genes associated with a phenotype with an undefined importance among the candidates in the list. Here, we validate the combination of GWAS, machine learning, and pathogenic bacterial population genomics as a novel scheme to identify SNPs and rank allelic variants to determine associations for accurate estimation of disease phenotype. This approach parsed a dataset of 1.2 million SNPs that resulted in a ranked importance of associated alleles of Campylobacter jejuni porA using multiple spatial locations over a 30-year period. We validated this approach using previously proven laboratory experimental alleles from an in vivo guinea pig abortion model. This approach, termed BioML, defined intestinal and extraintestinal groups that have differential allelic variants that cause abortion. Divergent variants containing indels that defeated gene callers were rescued using biological context and knowledge that resulted in defining rare and divergent variants that were maintained in the population over two continents and 30 years. This study defines the capability of machine learning coupled to GWAS and population genomics to simultaneously identify and rank alleles to define their role in abortion, and more broadly infectious disease.

Phylogenetic and Biogeographic Patterns of Vibrio parahaemolyticus Strains from North America Inferred from Whole-Genome Sequence Data

“…Pangenome analysis. The pangenome was determined as described previously by Bandoy and Weimer (56). Briefly, the genome sequences were assembled using Shovill (https://github.com/ tseemann/shovill), annotated using Prokka (57), and used as the data input for pangenome analysis using Roary (58).…”

“…Genome variation was determined using total genomic distance with a k-mer (31-mer) approach as a method to use the entire genome to determine relatedness between isolates as previously described (56). Population partitioning using nucleotide k-mers (PopPUNK) was used to determine related genomic clusters based on the whole-genome distance of the Vibrio isolates used in this study.…”

Dual-Organ Transcriptomic Analysis of Rainbow Trout Infected With Ichthyophthirius multifiliis Through Co-Expression and Machine Learning

“…Here, we use dual-organ RNA sequencing (RNA-seq) approaches to characterize the despotic temporal transcriptome in detail during hematopoietic stem and progenitor cell expansion in the rainbow trout. Tools based on machine learning technology and WGCNA can make high-dimensional data analysis simple and help to identify featured genes, hub genes, and/or key factors ( 23 , 27 ). If multi-dimensional transcriptomics results are interpreted through several well-known methods, the advantages of all methods can be maximized by compensating for the limitations of each.…”