ABSTRACT
SUMMARY: This work describes two novel workflows for variant calling that extend the widely used algorithms of Strelka2 and FreeBayes to call somatic mutations from multiple related tumour samples and one matched normal sample. We show that these workflows offer higher precision and recall than their single tumour-normal pair equivalents in both simulated and clinical sequencing data. AVAILABILITY AND IMPLEMENTATION: Source code freely available at the following link: https://atlassian.petermac.org.au/bitbucket/projects/DAW/repos/multisamplevariantcalling and executable through Janis (https://github.com/PMCC-BioinformaticsCore/janis) under the GPLv3 licence. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Subject(s)
Neoplasms , Software , Humans , Workflow , Algorithms , Neoplasms/geneticsABSTRACT
Muscle Invasive Bladder Cancer (MIBC) has a poor prognosis. Whilst patients can achieve a 6% improvement in overall survival with Neo-Adjuvant Chemotherapy (NAC), many do not respond. Body fluid mutant DNA (mutDNA) may allow non-invasive identification of treatment failure. We collected 248 liquid biopsy samples including plasma, cell pellet (UCP) and supernatant (USN) from spun urine, from 17 patients undergoing NAC. We assessed single nucleotide variants and copy number alterations in mutDNA using Tagged-Amplicon- and shallow Whole Genome- Sequencing. MutDNA was detected in 35.3%, 47.1% and 52.9% of pre-NAC plasma, UCP and USN samples respectively, and urine samples contained higher levels of mutDNA (p = <0.001). Longitudinal mutDNA demonstrated tumour evolution under the selective pressure of NAC e.g. in one case, urine analysis tracked two distinct clones with contrasting treatment sensitivity. Of note, persistence of mutDNA detection during NAC predicted disease recurrence (p = 0.003), emphasising its potential as an early biomarker for chemotherapy response.