3.1 Potential well-documented target driver genes from the CSCO guideline for NSCLC using NGS

According to the NSCLC CSCO 2022 guideline, detection of SNV/indel of EGFR, KRAS, HER2 and BRAF genes, CNV of MET and HER2, gene fusions of ALK, ROS1, RET and NTRK, and MET exon 14 skipping (METΔex14) would facilitate clinical application of target drugs for NSCLC patients based on these gene variations. DNA panel sequencing (769 genes) covering all aforementioned driver genes was used for the SNV/indel detection of EGFR, KRAS, HER2 and BRAF genes (Fig. 1A), and CNV of MET and HER2 genes (Fig. 1B). The number of SNV/indel of EGFR was 201 cases (52.1%), and the number of CNV of MET was 16 cases (4.1%) in the NSCLC cohort (Fig. 1A, B). Gene fusions of ALK, ROS1, RET, NTRK, and METΔex14 were detected by sequencing of DNA and RNA panels (Fig. 1C).

Fig. 1

Roles of NGS detection in diagnostic and treatment of NSCLC according to the Chinese Society of Clinical Oncology (CSCO) clinical guidelines. A SNV/indel mutations of EGFR, KRAS, HER2 and BRAF in NSCLC cases (n = 386) detected by DNA panel sequencing. B CNV mutations of MET and HER2 in NSCLC cases (n = 386) detected by DNA panel sequencing. C ALK, ROS1, RET and NTRK fusion and METΔex14 in NSCLC cases (n = 386) detected by DNA and RNA panel sequencing. D Percent of ALK, ROS1, RET and NTRK fusion and METΔex14 detected in NSCLC cases by DNA and RNA sequencing, respectively. E–F Gene fusion number E and fusion types F of ALK, ROS1, RET, NTRK and METΔex14 detected by DNA or RNA panel sequencing

Gene fusions were further explored using DNA panel or RNA panel sequencing in the NSCLC cohort. However, fisher's exact test showed that of the DNA and RNA panel sequencing was not significantly different in detection of ALK, ROS1, RET, NTRK, and METΔex14 genes (p > 0.05, Fig. 1D). The number of gene fusion containing ALK, ROS1, RET, NTRK or METΔex14 was comparable in the NSCLC cohort (Fig. 1E). 50% (26/52) gene fusion was simultaneously detected by both strategies and the remaining 50% was complementarily found by DNA or RNA panel sequencing (Fig. 1E). Surprisingly, detection rate of gene fusion types with ALK, ROS1, RET, NTRK or METΔex14 by DNA panel sequencing were higher than that using RNA panel (Fig. 1F). These data indicated that DNA sequencing combined with RNA sequencing could more comprehensively and reliably exhibit variations of genes in the NSCLC cohort.

3.2 Gene variation landscapes in the NSCLC cohort with DNA and RNA panel sequencing

We further displayed SNV/indel landscape of top 20 genes using DNA panel sequencing in NSCLC cases, which indicated that the SNV/indel frequencies of TP53 and EGFR were 60% and 55%, respectively (Fig. 2A). CNV landscapes of all genes in the NSCLC cases were showed in Fig. 2B. CNV rates of CDKN2A, myelocytomatosis oncogene (MYC), EGFR and cyclin-dependent kinases 4 (CDK4) genes in the NSCLC cohort were 15.9%, 12.4%, 8.2% and 7.4%, respectively. CDKN2A gene was loss of copy number, and MYC, EGFR and CDK4 genes were gain of copy number. Gene fusion landscapes was showed in Fig. 2C in the NSCLC cohort using the DNA and RNA panel sequencing. Echinoderm microtubule-associated protein-like 4 (EML4)-ALK fusion detected by NGS was most in all gene fusions and the rate was 8.2% in the NSCLC cohort. Although fusion detection rate of genes from the CSCO guideline with DNA and RNA panels was similar (Fig. 1E, F), the number and types of gene fusion detected using the RNA panel sequencing were more than those through the DNA panel sequencing (Fig. 2D, E). Thus, sequencing of the RNA panel greatly complemented deficiency of the DNA panel for gene fusion in the NSCLC cohort, which cooperatively revealed a more comprehensive gene variation feature in NSCLC.

Fig. 2

Landscape of SNV/indel, CNV and gene fusion in the NSCLC cohort. A SNV/indel landscape of top 20 genes using DNA panel sequencing in NSCLC cases (n = 365). Gene variations contained SNV, insertion, deletion and complex mutations. B CNV landscape using DNA panel sequencing in NSCLC cases (n = 172). There were 58 genes with CNV in the cohort. Gene copy number was showed using log2(CNV + 1). C Gene fusion using DNA and RNA panels sequencing in NSCLC cases (n = 386). Orange, blue and green presented fusion genes of RNA panel, DNA panel and common detection with RNA and DNA panels, respectively. D Venn diagram of the number of fusion genes detected by DNA panel and RNA panel in NSCLC cases (n = 386). E Venn diagram of the number of gene fusion types detected by DNA panel and RNA panel in NSCLC cases (n = 386)

3.3 Detection of rare gene fusion with DNA and RNA panel sequencing

By definition, fusion gene that was detected only once with DNA and/or RNA panels in the NSCLC cohort, was regarded as a rare gene fusion. As shown in Fig. 3A, there were 40 rare fusions identified in the cohort. Among them, 4 gene fusions, including F-box and leucine-rich repeat protein 20 (FBXL20)-EGFR, fibroblast growth factor receptor (FGFR4)-PYGO1, hedgehog-interacting protein 1(HIP1)-ALK and interferon regulatory factor-2-binding protein-2 (IRF2BP2)-NTRK1, were simultaneously discovered using DNA and RNA panel sequencing. 14 (35%) rare fusions were detected by the DNA panel and 22 (55%) rare fusions were detected by the RNA panel. The gene fusion with intergenic region was only detected by DNA panel, including intergenic (AKIRIN1, NDUFS5)-NTRK2, intergenic (ZNF318, ABCC10)-EZR, intergenic (POTEC, ANKRD30B)-PDGFB, intergenic (LINC00487, NRIR)-ALK and ALK-intergenic (ALK, YPEL5) (Fig. 3A). Gene fusion with driver genes such as TAF4-MET and SND1-MET detected by the RNA panel sequencing might provide a target-drug direction (Fig. 3B, C). In addition, the gene variants outside the spectrum of panels were not measured, which needs to be further assessed using whole genetic sequencing or transcriptome sequencing.

Fig. 3

Detection of rare gene fusion with sequencing of the DNA and RNA panels. A Intergenic(AKIRIN1,NDUFS5)-NTRK2 fusion was only detected by sequencing of DNA panel. B TAF4-MET fusion was only detected by sequencing of RNA panel. C SND1-MET fusion was only detected by sequencing of RNA panel

3.4 METΔex14 detection using DNA and RNA sequencing

In the NSCLC cohort, we found five patients with METΔex14 using DNA or RNA panel sequencing. DNA sequencing showed that 11 NSCLC cases with MET mutation and only one case was MET mutation at a c.3028G > A site, which predicted MET exon 14 skipping in mRNA processing. The METΔex14 was also confirmed using RNA panel sequencing (Fig. 4A). Four cases with METΔex14 were only found through RNA panel sequencing (Fig. 4B). Thus, METΔex14 was easier detected using RNA sequencing compared to DNA sequencing.

Fig. 4

METΔex14 detection using DNA and RNA sequencing. A Sequencing of DNA panel showed MET mutation at c.694_695insA (protein: p.S232Yfs*13) and sequencing of RNA panel showed METΔex14 in the patient. B Representative METΔex14 data was only detected with sequencing of RNA panel in the patient

3.5 Correlation of SNV/indel/CNV/fusion with clinical characteristics

We further analyzed correlation of SNV/indel/CNV/fusion with clinical characteristics in the NSCLC cohort. As shown in Fig. 5A–C, SNV/indel rates of TP53, EGFR and KRAS were significantly correlated with gender (Fig. 5A), smoking (Fig. 5B) and cancer subtype (Fig. 5C). Of course, some genes, such as EGFR were significantly different in different age (Figure S1A) and clinical stage (Figure S1B). CNV including cyclin D1 (CCND1) and FGF3/4/19 genes in male, smoking and LUSC was significantly enhanced than no-smoking and LUAD, respectively (Fig. 5D–F). CNV of CCND1 and FGF3/4/19 genes were dramatically different between stage II and III/IV patients (Fig. 5G). But there was no statistical correlation in CNV of many genes with age (Figure S1C). Fusion rates of most genes in different groups according to age (Figure S1D), gender (Figure S1E), smoking (Figure S1F), tumor subtype (Figure S1G) and clinical stage (Figure S1H) exhibited subtle difference in the NSCLC cases. Overall, SNV/indel and CNV of many genes were correlated with some clinicopathological features, such as gender, smoking, cancer subtypes and clinical stages in NSCLC patients.

Fig. 5

Correlation of gene SNV/indel/CNV/fusion and clinical characteristics in the NSCLC cohort. A SNV/indel rates of top 20 genes between male and female in NSCLC cases. Statistics based on the Fisher’s exact test. *p < 0.05, **p < 0.01. B SNV/indel rates of top 20 genes between smoking and non-smoking patients with NSCLC. Statistics based on the Fisher’s exact test. *p < 0.05, **p < 0.01. C SNV/indel rates of top 20 genes in patients with LUAD or LUSC. Statistics based on the Fisher’s exact test. *p < 0.05, **p < 0.01. D CNV difference analysis of NSCLC cases between male and female. Statistics based on the unpaired t test. *p < 0.05, **p < 0.01. E CNV difference analysis between smoking and non-smoking patients with NSCLC. Statistics based on the unpaired t test. *p < 0.05, **p < 0.01. F CNV difference analysis in patients with LUAD or LUSC. Statistics based on the unpaired t test. *p < 0.05, **p < 0.01. G Difference analysis of CNV in NSCLC patients with stage II, III and IV. Statistics based on the One-Way ANOVA test. *p < 0.05, **p < 0.01