To validate the performance of OncoChase-AS, we first tested the limit of mutation detection by using the Quantitative Multiplex Reference Standard, which contains 11 mutations across six cancer-related genes (Horizon Discovery, Cambridge, UK). Mutation detection was examined using the Acrometrix Oncology Hotspot Control (Thermo Fisher Scientific, Waltham, MA, USA), a mixture of more than 500 Catalogue of Somatic Mutations in Cancer (COSMIC) mutations across 53 genes. Four cell lines (HCT116, H1975, SW620, and HT29) were obtained from the Korean Cell Line Bank and used to assess the correlation with repeatability and reproducibility. To test the identification of CNAs, we used two primary tumors that are known to have CNAs, with the approval of the institutional review board of Catholic University of Korea. Genomic DNA was extracted from these samples using the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany).

We performed NGS analysis for the DNA samples with the OncoChase-AS cancer panel using an Ion S5 sequencer (Thermo Fisher Scientific). Raw sequence data were analyzed with the Torrent Suite (Thermo Fisher Scientific). The aligner for mapping reads to the reference genome is included in the Torrent Suite [10]. To call variants from the mapped sequence data, we used a plug-in in the Torrent Suite, Torrent Variant Caller (v5.2.2.41). In order to annotate the called variants with the queried knowledge database, we used ANNOVAR [11]. In order to verify variants that were not detected by the Torrent Variant Caller plug-in, we confirmed no-called variants using the Integrative Genomic Viewer (IGV) program [12] with binary alignment mapping (BAM) format files as raw sequencing data before calling variants [13].

To verify the limit of detection, we used the Quantitative Multiplex Reference Standard DNA (Horizon Discovery) sample, which contains 11 variants with variant allele frequencies (VAFs) ranging from 1% to 24.5%. We diluted the reference DNA from 100% to 10% and sequenced it with OncoChase-AS. The limit of detection and correlation between the expected and observed VAFs were calculated as described [14]. Correlation coefficient and linear regression analysis was performed using R.

To check the concordance of the detected variants, Acrometrix Oncology Hotspot Control DNA (Thermo Fisher Scientific) was used. NGS reactions were performed with the OncoChase-AS panel in triplicated runs. We annotated the variants using the ANNOVAR program [11]. The criteria for variant calling were as follows: exonic variants detected in more than 5% in the 1000 Genomes Project and the Exome Aggregation Consortium (ExAC) Project for East Asian and variants with >3% mutant allele frequency (MAF) [15, 16].

To check the repeatability and reproducibility of the detection of mutations, especially mutations with a low VAF, we prepared mixtures of the DNAs extracted from four cell lines (H1975, HCT116, HT29, and SW620). NGS experiments were performed in duplicate by different researchers independently.

DNA copy number profiling of the targeted sequencing data was performed using NEXUS Copy Number, v9.0 (BioDiscovery, El Segundo, CA). CNA regions were defined by a rank segmentation algorithm. In the rank segmentation algorithm, we set a threshold for segmentation of p = 1.0E–6. The thresholds for copy number gain and loss were 0.3 and −0.4 on a log₂ scale, respectively. The thresholds for amplification and homozygous deletion were 1.0 and −1.0 on a log₂ scale, respectively.

The custom NGS panel, named OncoChase-AS, was designed to detect 95 cancer-related genes with clinically important variants (Fig. 1). The 95 genes were selected based on the published literature and cancer databases, such as COSMIC [17] and GENIE [18]. Of the 95 target genes, 41 genes were selected for mutation screening, 10 genes were selected for CNA screening, and the other 44 genes were selected for screening of both mutations and CNAs (Fig. 1).

To determine the limit of mutation detection with OncoChase-AS, we used the Quantitative Multiplex Reference Standard DNA (Horizon Discovery), which contains 11 mutations with a 1% to 24.5% VAF across six cancer-related genes (BRAF, KIT, EGFR, KRAS, NRAS, and PIK3CA) (Table 1). When we performed NGS analysis with Onco-Chase-AS in duplicate, all of the expected mutations were consistently identified in both NGS analyses (Table 1). To assess the limit of mutation detection, we serially diluted the DNA sample from 50% to 10% and performed NGS analysis with each diluted DNA in duplicate. As a result, mutations with a very low MAF (<1% MAF) were successfully identified in most of the mutations (19/22, 86.4%). Mutations with a low MAF (1% to 3% MAF) were also successfully identified in most of the mutations (23/26, 88.5%). All mutations with >3% MAF were detected without exception (40/40, 100%) (Table 1). These results suggest that more than 86% of the mutations with <1% VAF can be identified by NGS analysis with OncoChase-AS. These results also suggest that any mutations with >3% frequency can be detected with this platform. When we calculated the correlation between the expected and observed VAFs by linear regression analysis, the R² value was 0.97 (Fig. 2). This result further supports the reliability of this system for variant identification.

To determine how completely OncoChase-AS NGS detected the expected mutations, we used Acrometrix Oncology Hotspot Control DNA (Thermo Fisher Scientific), which is a mixture of 55 tumors that contain more than 500 COSMIC mutations across 53 genes. Among the mutations in the Acrometrix Oncology Hotspot Control DNA, OncoChase-AS targeted 358 mutations. Therefore, in principle, 358 mutations are expected to detected, if OncoChase-AS NGS works perfectly. In the first analysis, 353 out of the 358 expected mutations were detected (98.6% concordance). In the second analysis, all 358 mutations were detected (100% concordance). In the third analysis, the concordance rate was 98.6% (353/358) (Table 2). These results indicate that OncoChase-AS is fairly sensitive and reliable in identifying mutations.

Cancer gene panels can be used for clinical tests, such as the identification of druggable mutations from tumor tissue or liquid biopsy, in addition to research. Therefore, the reproducibility of mutation detection is an important issue. To test the reproducibility of mutation detection, we performed OncoChase-AS NGS with four cell lines (H1975, HCT116, HT29, and SW620) harboring eight known mutations (Table 3). In this analysis, cell line mixture samples were sequenced in triplicate by two different researchers. All of the expected mutations in mixtures and the MAFs were measured in every run. With these results, we verified the repeatability and reproducibility of the Onco-Chase-AS panel. The average repeatability and reproducibility were 98% (ranging from 93% to 100%) and 98% (ranging from 95% to 100%), respectively (Table 3).

Chromosomal alteration is one of the most commonly occurring events during tumorigenesis. OncoChase-AS was designed to detect CNAs for 54 genes, including most of the clinically important CNAs, such as ERBB2 amplification and CDKN2A/B deletion. To test whether OncoChase-AS detected CNAs properly, we analyzed two primary tumors (brain cancer and sarcoma) with known CNAs: ERBB2 amplification in a brain cancer and CDKN2A deletion in a sarcoma. As a result, the expected CNAs were detected precisely by OncoChase-AS NGS analysis (Fig. 3). When we validated the target CNAs by genomic quantitative polymerase chain reaction (qPCR), the results were consistent with the NGS analysis results. Although we did not test all 54 target genes, the current results indicate that OncoChase-AS can robustly detect clinically important CNAs.

In conclusion, we have developed a cancer panel containing 95 cancer-related genes. In terms of the limit of detection, OncoChase-AS can detect more than 86% of target mutations with <1% MAF. In the case of those with >3% MAF, any mutation can be detected by NGS analysis with OncoChase-AS. When we applied this system for the analysis of Acrometrix Oncology Hotspot Control DNA, which contains more than 500 COSMIC mutations across 53 genes, 99% of the expected mutations were robustly detected. We also confirmed the very high reproducibility of the mutation detection in multiple independent analyses. When we examined CNAs, two expected CNAs related to tumorigenesis were successfully detected, and this result was confirmed by target-specific genomic qPCR. All of these results support the reliability and accuracy of our Onco-Chase-AS platform in detecting mutations. Therefore, this panel could be helpful for key mutation profiling research in solid tumors and clinical translational research, because this panel covers all 14 essential genes that have been suggested to be included in cancer panels for support from the national health insurance system in Korea. There are limitations of our panel. Due to the technical limitations of PCR amplification-based NGS, OncoChase-AS cannot detect repeat sequences with enough reliability and reproducibility, such as microsatellite instabilities and the promoter region of the TERT gene. For this purpose, a hybridization-based targeted sequencing panel would be more suitable.