Content uploaded by Bart C Weimer
Author content
All content in this area was uploaded by Bart C Weimer on Jul 16, 2019
Content may be subject to copyright.
Pangenome Wide Association Study Identifies Novel Allelic Variants Linked to
Livestock Abortion from Campylobacter jejuni Using Machine Learning
Abortive and non-abortive Campylobacter jejuni
in ruminants
1
2
Assembly
Genome Difference Matrix
b
d
Genome
Annotation
Pangenome
Bioinformatic Analysis Workflow
c
Identification &
AMR profiling
e
D. Bandoy1,2, A. Weiss1, N. Kong1, C. Huang1, B. Weimer1;!1University of California, Davis, Davis, CA; 100K Pathogen Genome Project,
2Department of Veterinary Paraclinical Sciences, College of Veterinary Medicine,University of the Philippines Los Baños
abortive
}
non-abortive outgroup
abortive
}
Figure 1. Genomic flux profile of abortive and non-abortive Campylobacter jejuni
f
Population
genome flux
Machine learning algorithm
Machine Learning modeling with population genomic flux
Figure 2. PorA Loop 4 model prediction of abortive and non-abortive Campylobacter jejuni
Isolation & Whole Genome Sequencing
}
Polygenomic population
associated with abortion
Accessory genome is
hypervariable
Single genome reference only
captures a single version of the
polygenome for the disease
a
b
3
Feature Extraction
(500,000 alleles)
c
Prediction Model
Generation
d
a
Prediction Model
Validation
e
f
Population Genome Flux Model Database
METHODS
RESULTS
CONCLUSION
core genes
variable genes
abortive
non-abortive
INTRO
The goal of the project is to determine novel biomarkers of abortion in ruminants due
to Campylobacter jejuni infection using predictive modeling of whole genome
sequencing and machine learning. Determining the genomic basis for phenotypes
is going to impact infectious disease surveillance, vaccine design and public health.
Here we applied a novel approach to population genomics of infectious disease using
machine learning.
Biomarkers associated with immunoevasion and virulence are highly
associated with abortion in Campylobacter jejuni.
Machine learning based genomic flux modeling is a novel approach
to population genomics of infectious disease, yielding insights and
novel biomarkers.
Bart C. Weimer,PhD bcweimer@ucdavis.edu
contact