Design of targeted next-generation sequencing (NGS) of circulating tumour DNA (ctDNA): towards overcoming DNA

Abstract

Background: Circulating tumour DNA being holds promise for addressing the challenge of developing NGS-based assays for somatic mutation detection. In this study we examined influence of amplicon-based panel design on assay accuracy considering non-random ctDNA fragmentation Methods: The sequencing was performed on ctDNA fragments extracted from plasma samples of 14 lung cancer patients. Paired sequencing libraries were prepared according to manufacturer’s protocol for Ion Ampliseq Cancer Hotspot Panel (ITCHP2) as well as custom panel (AODCP) for all samples. AODCP was designed to cover regions similar to the ones of ITCHP with the shorter amplicon length (median 137 vs 156). Nucleosome-guided ctDNA fragmentation patterns were analysed in publically available datasets (GSM1833219). To characterise fragmentation patterns intersecting ITCHP2 panel four descriptors were introduced: i) count of experimentally observed ctDNA fragments spanning the whole amplicon, ii) depth range, iii) depth change, iv) read depth shape (area between its linearization and itself) Results: Overall diagnostic sensitivity of NGS to detect EGFR mutations in ctDNA was 83%. Limiting analysis to the samples with plasma DNA amount of 80 ng and higher reduced the false negative rate from 17% to 0%. Across 51 either somatic or germline variants detected from libraries generated with varying amplicon length variant calling results were completely concordant. Across germline variations covered by both AODCP and ITCHP2, allele frequencies (af.) were equivalent (Wilcoxon signed rank test p-value = 0,88). Somatic mutations af. Pearson’s correlation coefficients was 0.88 (p-value = 0,44) for point mutations and 0.95 for EGFR exon 19 deletions (p-value = 0,53). Geometric mean ratio of the af. of the same EGFR exon 19 deletions detected by panels with different amplicon length was 1.16 (95% CI, 0.72-1.88; p-value >0.1). ctDNA fragmentation pattern analysis revealed averaged coefficient of variation of read depth shape, depth range and depth change of 390%, 68% and 38%. Descriptor equal-frequency discretisation (4 groups) resulted in both spanning fragments count and read depth coverage demonstrated correlation with coverage uniformity (ANOVA test p-value of 0,037 and 0,013). Descriptor quality estimation with RreliefF algorithm ranked spanning fragments count higher than GC content for predicting sequence coverage uniformity. Conclusions: Low cell-free DNA concentration remains the major limitation in liquid biopsy analysis and more stringent QC metrics in terms of DNA amount may therefore prevent senseless sequencing. Analytical and diagnostic sensitivity are irrelevant of the amplicon length at the specified range (140 vs 170 b.p.). Thus, designing highly multiplexed system where limiting amplicon length to 100bp is complicated, other parameters like melting temperature, cross homology, etc should be of higher priority than amplicon length. Finally, positioning of amplicons in accordance with the nucleosome placement in tissue of origin and genome-wide ctDNA fragmentation pattern thereof, may result in increased sequencing uniformity. Legal entity responsible for the study: Study protocol was approved by Atlas Biomed Internal Review Board Funding: NA Disclosure: Ivanov MV, E.Ignatova, A.Reznik, V. Mileyko: employment at Atlas OncoDiagnostics, all other authors have no conflicts of interest.