Leah Roberts

De novo identification of insertion sequences with De Bruijn graphs

Insertion sequences (IS) are small DNA elements that can replicate and move throughout a bacterial genome independently. This ability often results in their insertion upstream of or within genes, which consequently leads to large effects on gene expression within the bacteria. These changes in expression can affect a multitude of phenotypes, including resistance to antibiotics and virulence. As such, is it necessary that we characterise where these IS move to within the genome.

Unfortunately, as IS are small and repetitive throughout the genome, they cause issues for short-read de novo assemblers, most notably collapsed repeats. Because of this, it can be difficult to determine the exact location of IS throughout the genome.

This project aims to develop an end-to-end pipeline for de novo IS discovery using De Bruijn graphs, and quantify in a collection of bacterial genomes the effect of IS insertions on phenotype.

This project is suitable for an honours or Masters student. Some background in command line, HPC and python is highly desirable. In this project, you will learn about bacterial genomics and pipeline managers (e.g. Snakemake) in addition to bioinformatic tool development and testing. This project will be based at UQCCR (Herston Campus) and co-supervised by Dr Tom Stanton (Monash University).

Machine Learning to predict plasmids from bacterial isolates

Plasmids play a key role in gene exchange between bacteria and often carry gene conferring resistance to antibiotics and survival in hospital environments. However, they are difficult to fully characterise from short-read whole genome sequencing data alone. This is because plasmids are typically full of repeat sequences which can cause problems for short-reads assemblers. Long-read sequencing can solve this issue, however this technology is currently not routinely used in healthcare settings.

We have developed a plasmid network that allows users to predict the types of plasmids in their bacterial samples based on gene presence/absence. This project would build upon this work by creating a machine learning framework (Random Forest) that can predict plasmid presence from short-read bacteria.

This project is suitable for an honours or Masters student. Background in command line, HPC and python is highly desirable. This project will be based at UQCCR (Herston Campus) and co-supervised by Prof Zamin Iqbal (University of Bath, UK).

Pangenomes to predict bacterial transmission in healthcare settings

Predicting whether two bacterial isolates are the same (and thereby inferring if transmission has occurred) has traditionally been performed by identifying and counting single nucleotide variants (SNVs). To do this, a reference genome is usually selected, and isolate reads are mapped to the reference to identify SNVs in regions shared between all isolates. However, for large datasets of very diverse bacterial strains, a single reference genome is usually insufficient, as the shared regions between the strains becomes a very small proportion of the total genomic content.

We propose a novel method using pangenome reference graphs to better identify and discriminate transmission of bacterial pathogens. This project would start to build test datasets and develop novel workflows for predicting transmission from pangenome graphs.

This project is suitable for an honours, Masters, or PhD student. Background in command line, HPC and python is highly desirable. This project will be based at UQCCR (Herston Campus) and co-supervised by Dr Michael Hall (University of Melbourne).