Abstract The identification of mutations in circulating tumor DNA (ctDNA) can serve as a biomarker for non-invasive diagnosis and real time therapeutic monitoring. Although next generation sequencing (NGS) is a promising technology, the intrinsic low abundance of ctDNA in the high background of wild-type cell free DNA (cfDNA) makes the detection and quantification of such mutations in plasma a challenging task. This study aims to measure the performance of NGS analysis when applied to the identification of somatic mutations in plasma samples. Plasma samples from two healthy controls were used to isolate cfDNA. Constitutional genomic DNA from these controls was previously genotyped using the Human OmniExpress BeadChip. A custom multiplex PCR panel was designed to target 420 unique variants. To cover a broad range (0.5%-100%) of minor allele frequencies (MAF), as expected for somatic mutations in the plasma of cancer patients, the cfDNA samples from the two controls were mixed at five different rations (7:25, 1:6, 1:8, 1:20 and 1:100). Because sequence coverage is a key driver of accuracy, 5 different coverage profiles (100x, 500x, 1000x, 2000x and 5000x) were established for each of the cfDNA pools. On total 25 libraries were generated, pooled and sequenced with the Ion PGM system. Ion Reporter software was used for data analysis. The observed MAFs were compared with expected results for all the variants of each cfDNA pool. Median exon sequencing coverage of 133x, 650x, 1312x, 2628x and 5466x was obtained for each of the 5 cfDNA pools, with >99% of the amplicons covered. Overall base substitution and indel detection performance was high (>95% of the variants were detected across the different coverage levels). A direct correlation between expected and observed MAFs for the known variants was observed. For variants present at MAF ≥5%, a sensitivity of 91.4% and 100% was achieved at 100x and ≥500x coverage levels, respectively. For variants with 5% ≤ MAF ≤ 10% at 100x of coverage, the average quality score was <Q20. For variants occurring at low MAF (<5%) the detection sensitivity and the quality score dropped as coverage decreased. At 1000x coverage and above, more than 94% of the variants present at MAF <5% were successfully detected with an average quality score >Q30. Overall, positive predictive value remained high (>98%), with an average of 5 false-positive calls across the full coverage range. To validate our approach we analyzed 10 pairs of primary tumour/plasma samples. All cases carried known mutations of the EGFR gene detected during routine analysis. We were able to detect the corresponding mutation in the plasma samples in all cases. MAF in plasma samples ranged from 3% to 85%. Here we demonstrate that NGS can be used to detect mutations in cfDNA. Our approach with controlled data sets also enabled us to define allele frequency, allele depth, and allele quality score cutoffs for a reliable detection of low MAF variants in ctDNA. Citation Format: Ana Justino, Gabriela Fernandes, Ana Barroso, Barbara Parente, Venceslau Hespanhol, Jose C. Machado, Jose L. Costa. Next generation sequencing performance for the detection of mutations in plasma cell free DNA. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4930. doi:10.1158/1538-7445.AM2015-4930
Read full abstract