Project 467214

Cancer mortality rates can be greatly ameliorated by improving early diagnosis mechanisms and routinely screening the population. However, most tumour diagnosis methods are highly invasive and unfeasible for mass screening. The possibility of screening for cancer via liquid biopsy has gained recent popularity with the discovery of circulating tumour DNA present in blood plasma. Through Illumina sequencing of cell free DNA (cfDNA), multiple studies have been able to classify healthy and early-cancer patients. However, these approaches are expensive for routine screening and do not easily record methylation in the DNA, a key hallmark of cancer. We aim to improve upon these existing methods using Oxford Nanopore MinION sequencing. This device is small and relatively inexpensive, which is very suitable for a clinical setting. We plan to develop algorithms that extract key distinctions in cfDNA from tumour DNA and healthy DNA such as fragmentation and methylation patterns. In addition we will assess the reliability of mapping short Nanopore reads (~150bp) to the human reference genome and develop mappability masks to avoid erroneous alignments. We will improve upon our mapping accuracy by developing a novel alignment algorithm that maps reads to a pan-genome (A collection of reference genomes). Mapping reads to a single reference genome introduces reference bias and reduces racial equity in treatment efficacy. We aim to build a graph data structure to efficiently store and map to a collection of human reference genomes simultaneously. From using these data we aim to classify not only healthy and sick individuals, but also cancer types and tissue of origin. This protocol will greatly improve early cancer diagnosis rates when implemented in routine health screening check ups.