Real-world application of targeted next-generation sequencing for identifying molecular variants in Asian non-small-cell lung cancer

Abstract

Background

The advent of novel therapeutic agents has advanced biomarker characterization in non-small-cell lung cancer (NSCLC), driving increased adoption of next-generation sequencing (NGS) technologies for molecular testing. However, comprehensive data addressing the clinical utility of different NGS platforms for NSCLC remains limited.

Methods

This retrospective study analyzed real-world data from 478 Taiwanese NSCLC patients over five years, using the Oncomine Focus Assay (OFA) to assess genetic alterations. The evaluation focused on assay accuracy, limit of detection (LoD), sequencing performance, and the genetic landscape of NSCLC.

Results

The OFA achieved an NGS success rate of 80.5% (385/478), with tumor cell percentage, specimen source and FFPE block age identified as key factors affecting success. Quality metrics demonstrated robust sequencing performance, including 97.0 ± 9.6% on-target alignment, 94.7 ± 6.4% uniformity, and ≥ 500 × coverage for 98.0 ± 6.6% of amplicons. Among the 385 patients analyzed, 86.8% (334/385) were found to harbor pathogenic or likely pathogenic variants, of which 78.4% (262/334) were SNVs/Indels, 41.6% (139/334) were CNVs, 2.7% (9/334) were exon skipping alterations, and 10.2% (34/334) were gene fusions. Actionable driver mutations included EGFR mutations (46.2%, 178/385), KRAS mutations (9.4%, 36/385), ERBB2 mutations (6.8%, 26/385), ALK fusions (4.4%, 17/385), MET exon 14 skipping (2.3%, 9/385), BRAF mutations (2.3%, 9/385), ROS1 and RET fusions (1.8%, 7/385 each), and NTRK1 fusions (0.5%, 2/385). Notably, KRAS G12 C mutation was detected in 2.8% (11/385) of cases.

Conclusions

This study demonstrates the robust performance of the OFA in identifying clinically relevant genetic alterations in NSCLC. The findings support its clinical utility in precision oncology and provide valuable insights into the genetic landscape of Asian NSCLC, enhancing personalized treatment strategies for lung cancer patients.

Background

Lung cancer remains the leading cause of cancer-related deaths globally and is responsible for approximately 1.8 million fatalities each year [1]. Non-small cell lung cancer (NSCLC), accounting for approximately 85% of cases, is often linked to a poor prognosis, largely because most patients are diagnosed at unresectable advanced stages [2]. Traditional treatments for NSCLC, such as chemotherapy and radiation, are limited by their broad, non-specific effects, targeting both cancerous and healthy cells [3, 4]. In recent years, the emergence of molecular targeted therapies has considerably transformed the treatment paradigm for NSCLC, with high potential for improved patient outcomes [5].

Molecular targeted therapies are designed to inhibit specific proteins or genetic mutations that are essential for tumor growth and survival. Many of these therapies have been developed by identifying driver mutations — genetic changes that promote cancer cell proliferation. Mutations in genes, such as EGFR, ALK, ROS1, BRAF, HER2, and KRAS, have been shown to drive tumor progression in a subset of NSCLC patients; therapies targeting these alterations have been shown to provide substantial clinical benefits, including significant tumor shrinkage, prolonged progression-free survival, and, in some cases, improved overall survival [6, 7]. Thus, molecular profiling has become an essential component in the diagnosis and treatment planning of NSCLC, guiding the selection of therapies tailored to the unique genetic composition of a patient’s tumor [8].

Next-generation sequencing (NGS) has fundamentally transformed molecular diagnostics of NSCLC by facilitating comprehensive genetic profiling [9]. Unlike traditional methods, such as Sanger sequencing and polymerase chain reaction (PCR)-based assays, NGS enables the simultaneous detection of various genetic alterations within a single assay, providing a more exhaustive characterization of the molecular profile of tumors.

The application of NGS in molecular testing of NSCLC offers several significant advantages [10,11,12]. First, NGS can effectively identify a broad array of genetic alterations that are critical for informing treatment decisions. The generated comprehensive genetic profile of a patient’s tumor enables clinicians to pinpoint actionable mutations, facilitating the selection of treatment strategies personalized to the unique characteristics of a tumor. Second, NGS offers improved sensitivity for detecting low-frequency mutations. Tumors are inherently heterogeneous, and different cells may harbor distinct genetic alterations. The ability of NGS to identify these low-frequency mutations is essential for understanding the mechanisms of treatment resistance and tumor progression.

However, the clinical implementation of NGS in NSCLC presents several challenges. A primary concern is the need to balance comprehensive genetic profiling with the practical limitations associated with tissue availability. Lung cancer biopsies often yield limited tissue samples, especially in advanced disease stages, which requires careful optimization of specimen utilization to obtain the relevant genetic information without compromising result quality [13, 14]. Furthermore, the turnaround time for NGS testing can significantly impact clinical decision-making, as the processes of sequencing and data analysis may take several days to weeks, potentially delaying the initiation of appropriate therapies [15].

In May 2024, Taiwan began incorporating NGS testing into its national health insurance (NHI) coverage for cancer patients, marking a significant advancement in oncological care. This initiative spans various cancer types, including NSCLC. By shifting toward personalized medicine, it is expected to improve the clinical management of NSCLC—optimizing treatment outcomes, reducing the use of ineffective therapies, and ultimately enhancing patient care. With the integration of NGS into NSCLC management in Taiwan, identifying practical and feasible solutions for clinical use is essential to advancing precision oncology.

The Oncomine Focus Assay (OFA) is a targeted next-generation sequencing (NGS) panel designed to uncover actionable mutations in 52 oncology-related genes across various cancer types, including NSCLC. This assay analyzes both DNA and RNA from formalin-fixed, paraffin-embedded (FFPE) tissue samples, allowing for the simultaneous detection of gene mutations, insertions/deletions (Indels), gene fusions, and copy number variations. Moreover, by utilizing amplicon-based techniques, OFA is more compatible with small tissue samples compared to hybrid capture-based NGS assays, making it particularly well-suited for cases with limited biopsy materials [16].

In this study, we evaluated the clinical performance of the OFA in detecting genetic alterations in NSCLC over a five-year period in a Taiwanese patient population. Our aim was to assess the reliability and versatility of the OFA and to characterize the distribution of various genetic alterations among Asian NSCLC patients. The findings underscore the practical utility of the OFA in clinical settings, advancing precision oncology and supporting treatment decisions for NSCLC.

Methods

Study population and materials

This study was approved by the Institutional Review Board (IRB) of Taipei Veterans General Hospital. Sequencing data and quality metric reports were collected from 478 NSCLC samples submitted for the Oncomine Focus Assay (OFA) at Taipei Veterans General Hospital between 2019 and 2023. For assay validation, Horizon OncoSpan (Cat. No. HD827, Horizon Diagnostics, Waterbeach, Cambridge, UK), Horizon Structural Multiplex Reference Standard (Cat. No. HD753, Horizon Diagnostics, Waterbeach, Cambridge, UK), Seraseq CNV Mix Standard (Cat. Nos. 0710–0414 & 0710–0411, Seracare Life Sciences, Ozyme, Saint Cyr l’Ecole, France), and Seraseq Fusion RNA Mix v3 (Cat. No. 0710–0431, Seracare Life Sciences, Ozyme, Saint Cyr l’Ecole, France) were used.

Nucleic acid extraction

All formalin-fixed, paraffin-embedded (FFPE) tumor tissues and cytology samples prepared as paraffin-embedded cell blocks were reviewed by experienced pathologists to assess tissue quality and tumor content. Only samples with ≥ 20% tumor cells were considered acceptable for the OFA. However, in certain clinical scenarios, such as when patients experience rapid disease progression, cannot undergo repeat biopsies, or lack additional tumor tissue, suboptimal specimens may be used for NGS testing after discussion with the clinicians to provide critical molecular insights for timely treatment decisions. A designation of suboptimal specimens was included in the final reports for notification. The selected samples were deparaffinized, and DNA and RNA were extracted from the samples using the RecoverAll Total Nucleic Acid Isolation Kit (Thermo Fisher, Les Ulis, France) according to the manufacturer’s protocol. The extracted DNA and RNA were quantified using a Qubit 4.0 fluorometer with Qubit 1X dsDNA and Qubit RNA high-sensitivity assays (Invitrogen, Thermo Fisher, Waltham, USA).

Next-generation sequencing

The Ion Torrent Oncomine Focus Assay is a targeted next-generation sequencing (NGS) platform that analyzes DNA and RNA from 52 oncology-related genes, detecting single-nucleotide variants (SNVs), insertions and deletions (Indels), copy number variations (CNVs), and gene fusions within a single assay. The DNA panel detects hotspot mutations in genes such as AKT1, ALK, AR, BRAF, CDK4, CTNNB1, DDR2, EGFR, ERBB2, ERBB3, ERBB4, ESR1, FGFR2, FGFR3, GNA11, GNAQ, HRAS, IDH1, IDH2, JAK1, JAK2, JAK3, KIT, KRAS, MAP2 K1, MAP2 K2, MET, MTOR, NRAS, PDGFRA, PIK3 CA, RAF1, RET, ROS1, and SMO, as well as 19 CNVs, including ALK, AR, BRAF, CCND1, CDK4, CDK6, EGFR, ERBB2, FGFR1, FGFR2, FGFR3, FGFR4, KIT, RAS, MET, MYC, MYCN, PDGFRA, and PIK3 CA. The RNA panel identifies rearrangements in genes such as ALK, RET, ROS1, NTRK1, NTRK2, NTRK3, FGFR1, FGFR2, FGFR3, MET, BRAF, RAF1, ERG, ETV1, ETV4, ETV5, ABL1, AKT3, AXL, EGFR, ERBB2, PDGFRA, and PPARG. For RNA panel analysis, complementary DNA (cDNA) synthesis was performed using the SuperScript™ VILO™ cDNA Synthesis Kit (Thermo Fisher Scientific, Cat. No. 11754050). A total of 20 ng of DNA or RNA were utilized for library preparation, following the manufacturer’s protocol. The obtained DNA or RNA libraries were normalized to 60–80 pM and pooled for template preparation on the Ion Chef System (Thermo Fisher Scientific) using the Ion PGM Hi-Q Chef Kit. Sequencing was then performed on an Ion S5 XL instrument (Thermo Fisher Scientific). The quality control (QC) criteria for each sample included a mean depth greater than 500×, with at least 90% of the reads mapped within the target region and a minimum coverage uniformity of 80%.

Variant calling was performed using Ion Reporter software (version 5.16) with the manufacturer's default criteria. For DNA variant calling, SNVs were identified at a minimum variant allele frequency (VAF) of 4%, and Indels at a VAF of 7%. Hotspot genes were detected with a minimum VAF of 3%. CNVs were reported when the median absolute pairwise difference (MAPD) was below 0.5 and the amplification factor was greater than 4. For RNA variant calling, targeted fusions required a minimum of 20 read counts, while non-targeted fusions, such as EGFR variant III and MET exon 14 skipping, required at least 120 read counts. Variants meeting these thresholds were clinically annotated using the Oncomine Knowledgebase Reporter (Thermo Fisher Scientific).

Archer Fusionplex Lung Panel assay

The Archer FusionPlex Lung Panel assay is designed to detect key gene fusions in seven genes, including ALK, BRAF, FGFR, NRG1, NTRK, RET, and ROS1, as well as exon skipping events in EGFR vIII and MET exon 14, along with select point mutations in 14 critical lung cancer-associated genes. The assay uses a targeted enrichment method known as anchored multiplex PCR (AMP) chemistry to focus on regions of interest. Briefly, at least 100 ng of RNA isolated from FFPE tumor samples are reverse transcribed into cDNA. The synthesized double-stranded cDNA undergoes end repair, adenylation, and ligation with molecular barcodes and a universal primer site. The resulting cDNA is amplified using nested gene-specific primers (GSP1 and GSP2). GSP2 is tagged with a common sequencing adapter in preparation for sequencing. The use of gene-specific and universal primers enables detection of both known and novel gene fusion partners. Libraries are templated on Ion Sphere Particles by the Ion Chef and sequenced on the Ion S5 XL System. Data analysis is performed using Archer Analysis version 6.2.7. The samples must meet the following quality criteria: presequencing RNA Ct ≤ 31, library quantity > 30 pM, and an average of at least 10 unique RNA start sites per control GSP2. A sample is considered positive when the fusion breakpoint is supported by at least two unique reads.

Reverse transcriptase PCR for MET exon 14 skipping detection

The RNA isolated from the FFPE tumor samples was reverse transcribed into cDNA using the SuperScript™ VILO™ cDNA Synthesis Kit (Thermo Fisher, Cat No. 11754050). The reverse-transcription PCR procedure was as follows: preheating at 25 °C for 10 min, incubation at 42 °C for 60 min, and a final step at 85 °C for 5 min. Subsequently, 20 ng of cDNA were analyzed for MET exon 14 skipping using the AmpliTaq Gold® 360 DNA Polymerase Kit (Applied Biosystems). A primer set spanning the junction from exon 13 to exon 15, METex14 F1 (5’-TTGGGTTTTTCCTGTGGCTG- 3’) and METex14 R1 (5’-GGATACTGCACTTGTCGGCA- 3’), was used to assess the MET exon 14 skipping status. The presence of 18S rRNA was confirmed using primers 18S rRNA F1 (5’-GATGGGCGGCGGAAAATAG- 3’) and 18S rRNA R1 (5’-GCGTGGATTCTGCATAATGGT- 3’). The thermal cycling procedure was as follows: initial denaturation at 95 °C for 2 min, 45 cycles of denaturation at 95 °C for 20 s, annealing at 60 °C for 30 s, extension at 72 °C for 60 s, and a final extension at 72 °C for 5 min. The PCR products were separated by electrophoresis on a 2% agarose gel.

Statistical analysis

Categorical variables were summarized as frequencies and proportions. Comparisons between groups were conducted using the Chi-square test or Fisher's exact test, as appropriate. Continuous variables were presented as medians, and group comparisons for continuous variables were performed using Student's t-test. A p-value of < 0.05 was considered statistically significant. Statistical analyses were performed using GraphPad Prism v.8.0. Additionally, cBioPortal was employed to assess the percentage of alterations and create OncoPrints for visualizing these alterations.

Results

Assessment of the limit of detection

Various commercial reference standards were used to assess the limit of detection (LoD) of the OFA. The LOD for SNVs and Indels was determined using the OncoSpan DNA Reference Standard (HD827) and the Structural Multiplex Reference Standard (HD753) from Horizon Diagnostics (Cambridge, UK). HD827 contains 16 variants (2 Indels and 14 SNVs), while HD753 contains 4 variants (2 Indels and 2 SNVs). Of the 20 variants with VAFs ranging from 1 to 32%, 17 were successfully identified by the OFA. However, three variants with expected VAFs of 1%, 2%, and 3% were not detected (Supplemental Table 1). The LoD for CNVs was determined using the Seraseq CNV Mix Standard (Seracare Life Sciences, Milford, MA, USA), which includes the FGFR3, MYC, ERBB2, MYCN, EGFR, and MET genes. All six CNVs, each expected to have 5 copies, were successfully detected, with observed copy numbers ranging from 6.7 to 10.2 (Supplemental Table 2).

The LoD for gene fusion was determined using the Seraseq Fusion RNA Mix v3 (Seracare Life Sciences, Milford, MA, USA) containing 16 types of fusions. These fusions include EML4-ALK, KIF5B-RET, NCOA4-RET, CD74-ROS1, SLC34A-ROS1, TPM3-NTRK1, FGFR3-BAIAP2L1, PAX8-PPARG1, FGFR3-TACC3, ETV6-NTRK3, LMNA-NTRK1, SLC45A3-BRAF, TMPRSS2-ERG, EGFR-SEPT14, MET exon 14 skipping, and EGFR variant III. All 16 fusion transcripts were successfully detected by the OFA at the original concentration, as well as at the 5× and 50× dilutions. The lowest number of fusion transcripts detected was 11 copies (Supplemental Table 3). These results demonstrate that the OFA reliably detects SNVs and Indels with VAFs of 5% or greater, CNVs with at least 5 copies, and gene fusions with at least 11 copies (Table 1).

Table 1 Summary of Limit of Detection (LoD) for the OFA

Accuracy

To assess the detection accuracy of the OFA for SNVs, Indels, CNVs, and gene fusions, a series of samples with known variants were analyzed. These samples include 32 proficiency testing (PT) samples from the College of American Pathologists (CAP), which contains 57 SNVs, 17 indels, 5 fusions, and 6 CNVs, as well as 30 clinical FFPE samples with 7 SNVs, 16 fusions, and 2 CNVs. Only prevalidated variants were included for evaluation, and any newly identified or unvalidated variants were excluded. Of the 110 variants assessed, only one—an NTRK1-LMNA fusion involving exons 1–9 of LMNA fused in-frame to exons 12–16 of NTRK1—was misidentified because of being outside the assay range. The overall detection accuracy was 99% with 100% accuracy for SNVs, indels, and CNVs, and 95% for gene fusions (Table 2).

Table 2 Summary of accuracy of the OFA

Assessment of interpretation criteria for MET exon 14 skipping variants

During the initial clinical implementation of the OFA, one of the test cases was identified as having a MET exon 14 skipping variant by the Ion Reporter software, with a read count of 205—exceeding the manufacturer’s recommended threshold of 120 reads for this variant. Given the rarity but clinical significance of MET exon 14 skipping in NSCLC, we further validated the result using reverse transcription-polymerase chain reaction (RT-PCR), which returned a negative result. To evaluate the appropriate interpretation cutoff for MET exon 14 skipping in the Ion Reporter software for OFA, we analyzed 10 FFPE samples from 9 lung cancer cases and 1 thyroid cancer case, all tested positive for MET exon 14 skipping, with read counts ranging from 205 to 11,681. These samples were then tested using both the Archer FusionPlex and RT‒PCR assays. The Archer FusionPlex Lung Assay is a targeted NGS assay specifically designed to detect critical genetic alterations in lung cancer, including EGFR vIII and MET exon 14 skipping. This assay enables simultaneous detection of multiple fusions in a single sequencing run, even without prior knowledge of fusion partners or breakpoints.

The results obtained from the three methods (OFA, Archer FusionPlex Lung Assay, and RT‒PCR) are summarized in Table 3. Among the 10 samples tested, only 4 were confirmed positive for MET exon 14 skipping by the Archer FusionPlex Lung Assay and RT-PCR. The corresponding read counts from the OFA for these positive samples were 9,954, 11,681, 17,777, and 28,098, respectively. All samples with read counts below 2,500 were negative for MET exon 14 skipping, as confirmed by both the Archer FusionPlex Lung Assay and RT‒PCR. Based on these findings, we updated the interpretation criteria for MET exon 14 skipping in the Ion Reporter software used in our laboratory. Samples with read counts below 2,500 are classified as negative, those above 9,000 are classified as positive, and samples with read counts between 2,500 and 9,000 are considered equivocal, requiring confirmation by additional assays.

Table 3 Comparison of MET exon 14 skipping detection across different platforms

Sequencing performance

The quality metrics for the OFA were evaluated across 166 runs including 385 clinical specimens over a 51-month study period. Each run produced an average of 3.69 × 10⁷ reads (minimum: 1.06 × 10⁷ reads). After filtering out polyclonal ion sphere particles (ISP), primer dimers, low-quality reads, and other test sequences, 61.8 ± 6.2% (CV = 10.1%) of the amplicons were retained for bioinformatics analysis. The average number of aligned bases was 2.37 × 10⁹, with 2.11 × 10⁹ (89.0%) bases achieving a quality score of 20 (Q20). On average, 1.86 × 10⁶ reads were mapped to the hg19 reference genome per sample, with 97.0 ± 9.6% of the reads aligning to the target region. The mean read length was 123.8 bp (± 11.9 bp). The uniformity of base coverage—defined as the percentage of on-target bases covered by at least 20% of the mean depth—was 94.7 ± 6.4%. Among the amplicons, 98.0 ± 6.6% (CV = 6.7%) achieved at least 500× coverage, 97.5 ± 4.1% (CV = 4.2%) showed no strand bias, and 97.4 ± 5.7% (CV = 5.8%) demonstrated end‒to‒end sequencing. The sequencing performance of 269 amplicons indicated a wide average coverage depth, with a mean coverage of 10,204× and a median of 4,832× (Fig. 1). The lowest average coverage was recorded for the amplicon OCP1_DCUN1D1_9 at 1,266×, whereas the highest coverage was recorded for the amplicon CHP2_EGFR_3 at 26,374×. Approximately 96% of the amplicons achieved an average coverage of at least 2,000×. These results indicate robust sequencing performance for the majority of the targeted amplicons in the assay.

Fig. 1

Average depth of coverage across 269 amplicons in clinical samples analyzed with the OFA. The X-axis represents the different amplicons, which are arranged in ascending order of average coverage (from left to right). The lowest average coverage observed was 1,266×, with approximately 96% of amplicons achieving an average coverage of at least 2,000×. The data are presented as means ± SDs. The amplicon IDs are listed in Supplementary Table 4

Performance on clinical NSCLC samples

During the study period, a total of 478 FFPE samples from NSCLC patients were submitted for analysis by the OFA. The patient characteristics are summarized in Table 4. The median age was 64 years (range: 30–96 years), with 234 patients (49.0%) being male. The median of tumor cell percentage was 40% (range: 1–90%). Of the specimens, 64.9% (310/478) were biopsy samples, 9.0% (43/478) were cytology samples prepared as paraffin-embedded cell blocks, and 26.2% (125/478) were surgical resections. Among the biopsy samples, the most common sampling method was CT-guided biopsy (39.7%), followed by bronchoscopic biopsy (17.7%) and sonography-guided biopsy (17.4%). The primary tumor was the most common tissue source for the assay. Ninety-three samples were excluded from further analysis due to insufficient tumor tissue (46.2%, 43/93), low tumor cell percentage (40.8%, 38/93), poor library quantity (8.6%, 8/93), and poor DNA or RNA quality (4.4%, 4/93) (Table 5). Given that the time interval between FFPE block preparation and NGS evaluation (referred to as FFPE block age) is a critical preanalytical factor influencing NGS success rates, we analyzed the block age in 478 samples. Of these, 85.1% (407/478) had a block age of less than 1 year, 11.1% (53/478) were 1 to 2 years, 2.1% (10/478) were 2 to 3 years, and 1.7% (8/478) were over 3 years (Table 4). Statistical analysis indicated that tumor cell percentage, specimen source, and FFPE block age are major confounding factors affecting the success of NGS analysis (Table 4, p < 0.0001). The median turnaround time from assay request to final reporting was 9 days (range: 4–30 days).

Table 4 Patient characteristics and sample conditions in this study (n = 478)

Table 5 Factors contributing to OFA failure (n = 93)

Among the 385 samples successfully analyzed, 334 (86.8%, 334/385) were found to carry pathogenic or likely pathogenic mutations, with a total of 117 clinically relevant variant types identified. Among the 334 patients, 262 (78.4%, 262/334) had SNVs/Indels, 139 (41.6%, 139/334) had CNVs, 9 (2.7%, 9/334) had exon skipping alterations, and 34 (10.2%, 34/334) had gene fusions (Fig. 2A). The 117 clinically relevant variants included 86 SNVs/Indels, 22 CNVs, and 9 gene fusions. The prevalence of actionable driver mutations was 46.2% (178/385) for EGFR mutations, 9.4% (36/385) for KRAS mutations, 6.8% (26/385) for ERBB2 mutations, 4.4% (17/385) for ALK fusions, 2.3% (9/385) for MET exon 14 skipping and BRAF mutations, 1.8% (7/385) for ROS1 and RET fusions, and 0.5% (2/385) for NTRK1 fusions. Among the KRAS mutations, the G12 C mutation accounted for 30.5% (11/36) of the cases (Fig. 2B). Additionally, EGFR exon 20 insertions were detected in 3.6% (14/385) of total cases, representing 7.9% (14/178) of cases harboring EGFR mutations, with the p.Ala767_Val769 dup mutation being the most frequent subtype.

Fig. 2

Distribution of clinically relevant variants in NSCLC patients. A Proportion of NSCLC samples with different types of genetic alterations, including single nucleotide variants (SNVs) and insertions and deletions (Indels), copy number variations (CNVs), MET exon skipping alterations, and gene fusions. B Prevalence of actionable driver mutations in NSCLC

The analysis of 262 patients with SNVs and Indels revealed mutations in a range of critical oncogenes and tumor suppressor genes (Fig. 3A). The most frequently mutated gene was EGFR, present in 68% (178/262) of patients, followed by KRAS (14%, 36/262), ERBB2 (10%, 26/262), and PIK3CA (8%, 21/262). Other genes with notable mutation rates included CTNNB1 (6%, 17/262), JAK3 (3%, 8/262), and BRAF (3%, 9/262). Less frequent mutations were observed in genes such as ERBB3, CDK4, FGFR2, and HRAS, each affecting less than 1% of the patients. Figure 3B displays the VAF distributions for each pathogenic or likely pathogenic SNV and Indel. The results highlight the genetic diversity of NSCLC, with mutations primarily observed in EGFR and KRAS, and also imply that some pathogenic variants may exist as minor clones within the tumors.

Fig.3

Genomic landscape of SNVs and Indels in NSCLC patients. A Oncoprint diagram showing the distribution of SNVs and Indels in the NSCLC patient cohort. The X-axis represents individual patient samples, and the Y-axis lists the mutated genes. Green indicates missense mutations, blue indicates Indels, pink represents complex mutations, and gray denotes no alterations. B Variant allele frequency (VAF) distribution for clinically relevant SNVs and Indels in the patient cohort. The X-axis shows the gene variant names, and the Y-axis represents the allele frequencies, presented as means ± standard deviations (SDs)

An analysis of 139 patients with CNVs resulted in the identification of copy number gains across a variety of genes, as shown in Fig. 4A. The most frequent copy number gain was observed in EGFR, affecting 36% (50/139) of patients, followed by MYC (16%, 22/139), ERBB2 and MET (both 15%, 21/139). Other genes with notable gains included NF1 (14%, 20/139), AR (11%, 15/139), CDK4 (10%, 14/139) and CCND1 (6%, 8/139). In the analysis of 43 patients with structural variants, including gene fusions and exon skipping, the EML4-ALK fusion was the most frequently observed (40%, 17/43), followed by MET exon 14 skipping alterations (21%, 9/43) and the CD74-ROS1 fusion (12%, 5/43). Other notable fusions included KIF5B-RET (9%, 4/43) and CCDC6-RET (7%, 3/43). Less common fusions, such as TPM3-NTRK1, were identified in 5% (2/43) of patients. The rarest fusions, EZR-ROS1 and SLC34A2-ROS1, occurred in 2.3% (1/43) of patients each (Fig. 4B). Taken together, these findings demonstrate the complex genetic landscape of NSCLC, offering insights into potential biomarkers that may be pivotal for the development of targeted therapies.

Fig. 4

Genomic landscape of CNVs and gene fusions in NSCLC patients. A Oncoprint diagram showing the distribution of copy number variations (CNVs) in the NSCLC patient cohort. The X-axis represents individual patient samples, while the Y-axis lists the affected genes. Red indicates gene amplifications, and gray denotes no alterations. B Distribution of gene fusions in the NSCLC patient cohort. The X-axis represents individual patient samples, and the Y-axis shows the fusion variants. Purple represents structural variants, pink denotes exon skipping, and gray indicates no alterations

Discussion

The increasing effectiveness of molecular targeted therapies in the treatment of lung cancer has led to an increased demand for comprehensive NGS testing to inform personalized treatment strategies. In this study, we evaluated the performance and clinical relevance of the OFA for genetic profiling in Taiwanese NSCLC patients. Our findings demonstrate that OFA provides high sequencing quality and reliably detects genetic alterations, including clinically actionable variants that are crucial for guiding targeted therapy decisions. Our results also highlight that integrating NGS into clinical practice allows for the detection of rare or less common mutations that might be overlooked by traditional methods, as well as the identification of minor subclonal variants that could contribute to tumor progression or therapeutic resistance. Therefore, the OFA proves clinically valuable in addressing the genetic heterogeneity of NSCLC and supporting informed treatment decisions.

In this study, 86.8% of the 385 NSCLC patients were found to harbor pathogenic or likely pathogenic mutations using the OFA. Among these, single nucleotide variants (SNVs) and insertions/deletions (Indels) were detected in 78.4% of patients, while 41.6% showed copy number variations (CNVs). MET exon 14 skipping alterations were identified in 2.7% of samples, and gene fusions were observed in 10.2% (Fig. 2A). The results reveal a high prevalence of clinically actionable genetic alterations in Taiwanese NSCLC, emphasizing the urgent need to implement comprehensive genomic profiling for optimizing personalized treatment strategies.

The use of NGS in oncology testing has been shown to improve patient outcomes by facilitating personalized treatment plans that target specific genetic alterations. A study conducted in the United States found a significant increase in next-generation sequencing (NGS)-based biomarker testing, with the proportion of patients tested rising from 28.3% in 2015 to 68.1% in 2020. The study also highlighted the clinical impact of targeted therapies, showing that patients with positive driver gene alterations who received targeted treatments had a median overall survival of 24.3 months, compared to 15.2 months for those with driver alterations who did not receive targeted therapy and 11.0 months for those without driver gene alterations [17]. However, NGS-based multigene testing remains underutilized in many Asian countries, including China, Malaysia, and Thailand, primarily due to the lack of reimbursement for genetic tests beyond EGFR and ALK. Notably, starting in May 2024, Taiwan's Ministry of Health and Welfare has approved national health insurance coverage for NGS-based testing in advanced NSCLC cases negative for EGFR mutations, marking a significant step forward in integrating comprehensive genomic profiling into clinical practice.

Sequencing quality is a key determinant in the successful application of NGS for cancer genomics, as it ensures robust variant detection and reproducibility, enabling the reliable identification of low-frequency or subclonal variants with significant clinical implications, particularly in tumor evolution and drug resistance. In this study, we demonstrated the robust sequencing performance of the OFA for oncology testing, with key findings including: 1) an average of 3.69 × 10⁷ reads per run, ensuring comprehensive assay coverage; 2) 97.0 ± 9.6% of reads aligning to the target region, reflecting precise targeting of clinically relevant genomic regions; and 3) high base coverage uniformity (94.7 ± 6.4%) and deep sequencing depth (mean coverage of 10,204×), supporting reliable detection of variants, even for low-frequency mutations. However, in our clinical practice, when sequencing quality did not fully meet acceptable criteria, variants with potential clinical significance were still reported if verified using alternative methods such as Sanger sequencing or qPCR, with comments on suboptimal sequencing quality included in the final reports. This approach allowed us to maximize the utility of sequencing data, particularly in cases with limited tissue samples, ensuring the identification and reporting of clinically relevant variants for clinical action.

Previous studies analyzing the oncogenic driver mutation profiles of Japanese NSCLC using the Oncomine Dx Target Test CDx (Oncomine DxTT) have reported the frequency of EGFR mutations ranging from 31.0% to 41.0%, KRAS G12 C mutation from 2.3% to 6.8%, BRAF mutations from 1.9% to 2.3%, ALK fusions from 2.0% to 3.4%, RET fusions from 0.0% to 2.3%, ROS1 fusions from 0.0% to 2.3%, NTRK fusions from 0.0% to 0.4%, MET exon 14 skipping from 3.0% to 3.4%, and ERBB2 mutations from 0.0% to 3.0% [18,19,20,21]. In comparison, our study, utilizing the Oncomine Focus Assay (OFA), identified a higher prevalence of EGFR mutations (46.2%) and ERBB2 mutations (6.8%) than previously reported. Several factors may contribute to these differences, including variations in assay design and performance, patient selection criteria, sample types, and other methodological specifications.

Notably, in those studies, compared to qPCR-based assays, some discordant results of EGFR mutations were observed, primarily due to the missed detection of exon 19 deletions by Oncomine DxTT. However, in the current study, we could not assess false-negative rates due to the lack of comparative data from other assays. Discrepancies between NGS and qPCR-based methods may arise from variations in the limit of detection (LOD), target coverage, or other assay-specific technical factors. Additionally, false negatives in NGS assays can result from bioinformatic challenges, particularly during the variant calling process. Studies have demonstrated that certain algorithms may have difficulty detecting complex EGFR mutations, including deletions, insertions, and duplications. [22, 23]. These bioinformatic challenges have gradually improved with ongoing software development and updates. During our laboratory’s initial implementation of the OFA, we occasionally observed missed calls for clinically relevant mutations, such as EGFR exon 19 deletions and ERBB2 exon 20 insertions, when using Ion Reporter software version 5.10. However, this issue was significantly mitigated after upgrading to version 5.16, emphasizing the importance of both regular software updates and thorough validation of bioinformatic pipelines to ensure the accurate and reliable detection of key oncogenic variants.

Implementation of NGS testing in clinical laboratories often faces the challenge of limited bioinformatics expertise, leading to reliance on vendor-provided analysis software [24]. Although these platforms offer standardized pipelines and interpretation criteria, they may not fully address the complexities of clinical applications. In this study, a commercial bioinformatics pipeline was used for NGS data analysis, with Torrent Suite Software automating primary analysis, data transfer, and secondary analysis in Ion Reporter Software. The default variant calling threshold for MET exon 14 skipping in Ion Reporter Software was 120 read counts. Given the rarity of this mutation, we initially verified it using alternative methods, including Archer FusionPlex assays and RT-PCR. Our evaluation showed that cases with read counts above 9,000 were consistently positive, those below 2,500 were reliably negative, and cases within the 2,500–9,000 range were classified as equivocal, requiring further confirmation (Table 3). By optimizing the variant calling threshold, we identified MET exon 14 skipping in 2.3% (9/385) of our NSCLC cohort, a detection rate consistent with the previously reported 1.7% prevalence in Asian NSCLC populations [25]. Notably, a previous study also reported false-positive results for MET exon 14 skipping detected by OFA, particularly in cases with low read counts [26]. In that study, 11 NSCLC cases with this alteration identified by OFA were analyzed, revealing that two cases with high read counts (≥ 2,540) were consistently confirmed positive across platforms, whereas cases with lower read counts (179–612) tested negative. These findings highlight the importance for optimized validation or verification of analysis software to ensure accurate identification of clinically significant variants, preventing misdiagnosis or inappropriate treatment decisions.

Tissue availability is another challenge for the implementation of NGS-based testing in NSCLC. Compared to traditional molecular assays, NGS-based testing demands higher specimen quality and quantity. Inadequate tissue samples can lead to suboptimal sequencing quality or incomplete genetic data, potentially compromising the accuracy of test results [27]. Specimens obtained from surgical resection are typically preferred for genetic testing due to their larger tissue quantities. However, in advanced NSCLC, tumors are often unresectable, and only tumor biopsy specimens are available for genetic testing, challenging comprehensive genomic profiling. In this study, the sequencing success rate of the OFA in our laboratory was 80.5% (385/478), with insufficient tumor tissue, low tumor cell percentage being the primary causes of failure (Table 5). Specimen source also played a key role in NGS success, with tumor biopsy specimens accounting for approximately 90% of the failure cases (Table 4). This finding aligns with a previous study in a Japanese population, which reported an NGS success rate of 80.1% in NSCLC [28]. Improving tissue acquisition strategies or combining tissue and liquid biopsy approaches could help address tissue availability challenges and enhance the accessibility and effectiveness of NGS-based testing in clinical practice.

Although FFPE tissues are commonly used in molecular oncology testing due to their routine collection in clinical practice for histopathological evaluation and their capability for long-term storage at room temperature, FFPE block age is a critical preanalytical factor that can affect nucleic acid quality [29]. Prolonged storage has been linked to decreased DNA and RNA extraction yields, which may impact subsequent molecular analyses [30, 31]. Several studies have examined the effects of FFPE storage on NGS performance. One study reported a continuous decline in NGS success rates over time, with success rates dropping to 50% after four years of storage [32]. Another study found that older specimens had significantly lower target region coverage and reduced average read depth, ultimately compromising the reliability of NGS assays [33]. In this study, we analyzed the block age of 478 samples and found a significant correlation between FFPE block age and NGS success rate (Table 4). Among the NGS success group, 89.6% (345/385) of samples had a block age of less than 1 year, 7.3% (28/385) were 1 to 2 years, 1.8% (7/385) were 2 to 3 years, and 1.3% (5/385) were over 3 years. In contrast, in the NGS failure group, 66.7% (62/93) of samples had a block age of less than 1 year, 26.9% (25/93) were 1 to 2 years, 3.2% (3/93) were between 2 to 3 years, and 3.2% (3/93) were over 3 years. These findings emphasize the importance of timely NGS testing following tissue preservation, as fresher samples are significantly more likely to yield successful results.

The present study has several limitations. First, as a single-institution study, the findings may not fully reflect variations in clinical practices or testing protocols across different health care settings. Future multicenter research with larger, diverse cohorts would provide a broader perspective and strengthen the conclusions. Second, the NGS data analyzed in this study were obtained from NSCLC samples submitted for the OFA during the study period, without considering the clinical status of the patients at the time of testing. Differences in tumor evolution, clonal selection, or treatment status could influence the genomic landscape, potentially affecting the interpretation of the findings. Third, patient selection bias may have influenced the prevalence and distribution of genetic alterations identified in this study. One possible source of selection bias is that, since EGFR and ALK testing have been separately reimbursed by Taiwan’s National Health Insurance for advanced NSCLC for years, the patients included in this study may have been those who were tested negative for these mutations or had developed resistance to standard treatments, and were therefore seeking alternative or additional therapeutic options. Furthermore, it is well-established that the genetic landscape of NSCLC varies across the major histological subtypes. Specifically, adenocarcinoma is more commonly associated with actionable driver mutations such as EGFR, KRAS, ALK, and ROS1 alterations. These potential biases may limit the relevance of our findings of mutation distribution to the broader NSCLC population. Finally, the OFA uses an amplicon-based approach to detect gene fusions, which is limited to known variants with specifically designed primers and cannot detect novel fusions. Previous studies have also suggested that amplicon-based assays may have lower sensitivity for detecting gene fusions compared to other platforms [34]. Future research should address this limitation to better evaluate the performance of the assay in detecting both known and novel gene fusions.

Conclusions

In Taiwan, NGS testing is classified as a laboratory-developed test (LDT) rather than an in vitro diagnostic (IVD) test, allowing greater flexibility for assay development and application. This study demonstrates the robust performance of the OFA in identifying clinically relevant genetic alterations in NSCLC, establishing it as a reliable NGS-based testing platform for clinical use. Through the analysis of NGS data from 385 Asian NSCLC samples, we gain a deeper understanding of the genetic landscape of NSCLC and provide valuable insights into the effective use and optimization of NGS in clinical practice. These findings can be directly applied to support more personalized treatment approaches for lung cancer patients worldwide.