Use of novel computational filtering to reduce background next-generation sequencing errors in circulating tumour DNA of metastatic colorectal cancer patients

Abstract

Next-generation sequencing (NGS) technology is emerging as a major technique for genotyping circulating cell-free DNA (cfDNA) and for patient monitoring. However, Results of NGS is subject to numerous errors. In this study, we isolated circulating cfDNA and genomic DNA from 39 available tumours from 54 patients with advanced colorectal cancer (CRC). Deep targeted sequencing was performed for a panel of 10 genes that are recurrently mutated in CRC. To reduce sequencing error, we devised a de-noising procedure and calculated the concordance of somatic variants between cfDNA and tumour tissue sequencing data. The sensitivity, specificity, and accuracy for somatic alterations in the 10 genes were increased from 84.5%, 74.6%, and 76.9% to 87.3%, 92.0%, and 91.1%, respectively, after de-noising. This approach improved the detection of somatic alterations in advanced CRC cfDNA. We could selectively detect clinically important somatic alterations for variant allele frequencies of 0.27%–79.42%. Patients with high cfDNA concentrations had more detectable somatic mutant fragments and larger liver metastatic lesions than patients with lower concentrations. These results demonstrate the suitability of de-noised deep targeted sequencing for cfDNA genotyping, and provide insights into strategies for monitoring metastatic lesions in patients with advanced CRC.