Somatic mutations, which accumulate in cancer cells, are caused by an admixture of endogenous and exogenous processes. These somatic mutations are readily detectable in cell free circulating tumour DNA (ctDNA), which is released by cancer cells into the blood stream, and can be used to detect ctDNA for cancer diagnosis and monitoring.
Here we present MisMatchFinder an unbiased method to infer active mutational processes from low coverage whole genome sequencing of either cell free or tissue DNA using a read-centric approach. Every read is individually compared to the reference genome, which allows our method to function without a matched normal and with very sparse data. Multiple steps of error suppression like read collapsing are used to reduce the impact of sequencing errors and low tumour purity. We observed a high concordance with the current gold standard analysis in multiple different cancer types (Breast, Melanoma, Lung, MDS) and data sets.
Our new method enables ctDNA serial tracking of cancer progression from a minimally invasive blood sample at a much lower cost than current approaches and can be used to distinguish healthy individuals from cancer patients with high accuracy. Further development of this method may facilitate large scale cancer screening and allow the assessment of cancer risk.