PDF
Citation
Print
Introduction
Cancer constitutes a substantial clinical burden worldwide, accompanied by the considerable cost of oncological therapies. In 2020, a staggering 19.3 million new cancer cases (18.1 million excluding nonmelanoma skin cancer) and nearly 10.0 million cancer-related deaths (9.9 million excluding nonmelanoma skin cancer) were reported globally [1]. In the United States alone, it was estimated that in 2023, there would be ~2 million new cancer diagnoses and a projected 610K cancer-related deaths [2]. This escalating prevalence of cancer underscores an expanding, unmet demand for genomic testing—a pivotal tool empowering healthcare providers to adopt a more precise and tailored approach to patient care.
In the pursuit of personalized medicine within oncology, next-generation sequencing (NGS) has emerged as a valuable clinical asset. Recent decades of sequencing endeavors have unequivocally established cancer as predominantly driven by genetic factors [3]. The disease manifests through a complex interplay of inherited germline mutations and acquired somatic mutations, which can confer a survival advantage while evading immune surveillance [4]. Through harnessing genomic and molecular profiling technologies to customize treatments based on a patient’s specific tumor alterations, personalized medicine has demonstrated its potential [5-7]. It enhances survival rates and overall quality of life and yields favorable economic outcomes compared to single-gene tests [8,9]. Despite these promising advantages, the adoption of tumor profiling remains limited, with only a fraction of eligible patients benefiting from genomic testing [8,10]. Perhaps this underutilization is linked to the existing mainstream tests that often fall short of comprehensively covering all clinically relevant cancer-related alterations. Conventional targeted panel sequencing (TPS) approaches, for instance, require some prior diagnostic hypotheses driven by previous clinical and/or genomic data, limiting their clinical utility [4,11]. Additionally, due to validation cycles required for updating panel tests, they may struggle to address emerging biomarkers promptly [10]. These constraints underscore the pressing need for a comprehensive examination of a patient’s genome.
The rapid evolution of sequencing technologies, coupled with substantial cost reductions, has ushered in the era of comprehensive genomic profiling (CGP) through whole-genome sequencing (WGS). WGS of tumor and matched normal tissues offers a comprehensive view of germline and somatic variants, including point mutations, copy number alterations, and structural rearrangements, delivering invaluable insights for precision oncology and personalized treatment strategies [12]. The comprehensive catalog of mutations further allows more accurate quality control of genome sequences and investigating patterns of mutations, known as “mutational signatures” [13]. As technology advances and costs decline, integrating WGS into routine comprehensive genomic profiling becomes increasingly indispensable to unlock the full potential of genomic medicine in cancer. Nonetheless, concerns have arisen regarding WGS’s sensitivity in detecting driver variants due to lower sequencing depth. Challenges also arise from low tumor cell fraction (TCF) and suboptimal DNA quality from formalin-fixed paraffin-embedded (FFPE) tissue, potentially leading to the oversight of low allelic fraction variants by chance [14].
In response to these challenges, a forward-thinking approach has been introduced: target-enhanced whole-genome sequencing (TE-WGS). Here, on top of the WGS backbone (40×), > 500 top key biomarker genes are more deeply explored by enrichments using pools of WGS libraries (Fig. 1). This unique amalgamation ensures greater confidence that even critical biomarkers with low variant allele fractions (VAFs) will not go undetected. This report presents the findings of the TE-WGS approach to cancer genomic profiling in comparison with TruSight Oncology 500 (TSO500) in real-world clinical settings.
Materials and Methods
1. Study design
This prospective observational study included adult participants with diverse cancer types and stages, conducted at Ajou University Hospital from September 2022 to March 2023 (registered to Clinical Research Information Service No. KCT0007707). Each sample underwent diagnosis and tumor content assessment by a pathologist. Routine cancer molecular profiling was performed on all patients using TruSight Oncology 500 (TSO500) (Illumina Inc.). Patients with residual tumor DNA after TSO500 testing provided peripheral blood samples for DNA extraction, enabling additional tumor-normal testing using TE-WGS conducted by CancerVision, a proprietary CGP product developed by Inocras Inc. [15]. A total of 49 tumor-normal pairs were included in the study analysis (Fig. 2).
2. DNA extraction
FFPE tumor DNA extraction was performed by the pathology laboratory at Ajou University Hospital using AllPrep DNA/RNA FFPE kit (Qiagen). Peripheral blood DNAs were extracted using Allprep DNA/RNA kits (Qiagen) by Inocras Inc.
3. Targeted panel
Illumina’s TSO500 NGS library preparation adhered to the manufacturer’s instructions, with a DNA target size of 1.28 Mb to enrich 523 genes. Sequencing was executed on the NextSeq 550 platform (Illumina Inc.), and data analysis occurred via TruSight Oncology 500 Local App (pipeline ver. 2.2, Illumina Inc.) on an on-premise server at Ajou University Hospital, utilizing the GRCh37 reference genome. Reported findings encompassed single nucleotide variants (SNVs) and insertions and deletions with a VAF ≥ 5%, copy number variations (CNVs) with ≥ 4 (gain) and < 1 (loss), and translocations supported by ≥ 5 translocation reads. Tier 1 or 2 SNVs and indels with VAF below 5% were reported if they possessed significant clinical implications or in the context of chemotherapy where the TCF might be diminished.
Tumor mutation burden (TMB) is the number of eligible somatic mutations per Mb. The eligible SNVs and indels need to meet the following criteria: (1) in coding regions, (2) in high-confidence regions, (3) with ≥ 50× coverage, and (4) with ≥ 5% VAF. For microsatellite instability (MSI), TSO500 analyzed 130 MSI marker sites to calculate a quantitative score (see Illumina’s application note 1170-2018-009-B).
4. Target-enhanced whole-genome sequencing
Residual tumor DNA from TSO500 testing by Ajou University Hospital and newly extracted peripheral blood DNA from Inocras Inc. were utilized to prepare paired tumor-normal sequencing libraries. Genome sequencing, analysis, and interpretation for TE-WGS were performed using the CancerVision system [15]. DNA libraries were created with TruSeq Nano Library Prep Kits (Illumina Inc.) and sequenced on the Illumina NovaSeq6000 platform (Illumina Inc.) with average depths of coverage of 40x for tumor samples and 20× for blood samples for WGS.
To enrich relevant targets, we selected 526 genes, which included all TSO500 genes along with SLC7A8, STT3A, and ZNF2, based on their established associations with the specific cancer type and their potential to guide clinical decision-making (clinical actionability). We then created a target bed file encompassing the coding exons of all selected genes and the promoter regions of TERT, resulting in a total of 6,746 targets with a combined size of 2.76 Mb. xGen Custom Hybridization Probes (IDT, Inc.) were designed based on this target bed file. Tumor DNA libraries were enriched using these probes, followed by sequencing on the Illumina NovaSeq6000 platform to achieve an average depth of coverage of 500×.
5. Bioinformatic analysis TE-WGS data
The obtained sequences were aligned to the human reference genome (GRCh38) using the BWA-MEM v1.0.4 [16]. Polymerase chain reaction duplicates were removed using SAMBLASTER v0.1.26 [17]. Germline small variant calling was initially performed using Inocras’ Clinical Laboratory Improvements Act (CLIA)–validated WGS analysis pipeline, which leverages both HaplotypeCaller2–GATK v4.2.0.0 [18] and Strelka2 v2.9.10 [19]. Somatic mutation calling was performed by germline subtraction from tumor mutations utilizing tools’ native function or in-house scripts. Somatic small variant calling was performed using Strelka2 v2.9.10 [19] and Mutect2–GATK v4.2.0.0 [20], and the variants were combined. Structural variations were identified using Manta v1.6.0 [21]. Identified germline and somatic variants were annotated using VEP release 107 [22] and manually inspected and curated using Inocras’ proprietary genome browser. Sequenza-based analysis provided estimates of tumor characteristics, including TCF, tumor cell ploidy, and segmented copy number alteration profiles [23].
Point mutation calling employed stratified cutoffs based on mutation type (hotspot vs. non-hotspot) and sequencing platform (WGS vs. TPS). Hotspots were defined according to loci described in the literature [24]. For hotspot mutations, a minimum VAF of 1% was required in tumor samples analyzed by TPS with at least 100× coverage depth. If TPS depth fell below 100×, the cutoff for WGS applied, requiring both VAF ≥ 1% and at least two variant allele reads. Non-hotspot mutations had stricter cutoffs. In WGS data, they needed VAF ≥ 5% and at least three variant allele reads. If WGS depth was sufficient (≥ 100×), the WGS cutoff mirrored TPS, requiring VAF ≥ 2.5%. However, for shallow TPS (depth < 100×), the non-hotspot cutoff reverted to the more stringent WGS criteria. These cutoffs balanced sensitivity for capturing genuine mutations with specificity to minimize false positives. CNV detection prioritized segmental alterations over arm-level changes. This ensures a high degree of confidence in identifying true amplification events. Biallelic deletions, where both copies of a gene are lost, were identified when the major and minor alleles’ copy numbers were equal to zero.
The TMB was calculated as the sum of SNVs and indels divided by an effective genome size of 2,862,010,428 bp [25]. MSI was assessed using a modified algorithm based on MSIsenssor [26], and the homologous recombination deficiency (HRD) score was estimated using a refined version of HRDetect, optimized for improved performance in FFPE tissue [27].
Results
1. Demographics
A total of 56 patients diagnosed with solid tumors between September 2022 and March 2023 were initially enrolled in the study. However, seven patients were subsequently excluded from future analysis due to insufficient DNA samples. Thus, a final cohort of 49 patients with available tumor-normal pairs was included in the subsequent analysis (Fig. 2). Among the participants, 31 (63%) were male, while the remaining 18 (37%) were female. The cohort’s median age was 59 years (range, 22 to 86 years). The cohort encompassed diverse tumor types, as detailed in Table 1. Briefly, these tumor types included breast (n=4), colorectal (n=7), pancreatic (n=11), stomach (n=8), and lung cancers (n=4).
2. Coverage depth
The mean depth of coverage for TSO500 was 830.1× (range, 378.5 to 1,035.1). In the TE-WGS approach, cancer whole-genome was sequenced to the mean depth of ~45× (range, 28.4 to 68.5) with enrichment for target genes to 359.8× (range, 84.4 to 896.5). Notably, 93.12% of the targeted regions achieved a sequence depth of 50× in the TE-WGS approach, whereas this value was 98.28% for TSO500 (Fig. 3A).
3. Variants detection
In order to evaluate the TE-WGS performance in variant detection of clinically actionable driver genes, genomic findings from the TSO500 were first considered. In total, 498 unique variants were identified across all samples in TSO500, with complete detection in the TE-WGS methods, accounting for the 100% sensitivity (498/498). The data demonstrated a high correlation of detection of VAF between the TSO500 and TE-WGS methodologies (r=0.978) (Fig. 3B).
Furthermore, among the pool of 498 commonly detected variants identified by both approaches, a notable proportion of 223 variants (44.8%) were discerned within peripheral blood samples through the CLIA-validated WGS (analytical sensitivity for SNV and indel is 99.8% and 99.2%, respectively, and positive predictive value for SNV and indel is 99.3% and 98.7%, respectively, by a separate validation study) [15], suggestive of constitutional variants inherent to the germline. Conversely, the remaining subset of 275 variants (55.2%) were not detected in peripheral blood through the blood DNA WGS, signifying them as bona fide somatic variants of the tumor tissue. Within the tumor TE-WGS data, the VAFs exhibited a spectrum of 5.0%-90.7% for constitutional variants and 2.1%-91.0% for somatic variants, with a substantial level of overlap. This underscores that the VAF may not be relied upon as a definitive criterion to distinguish germline and somatic variants (Fig. 3B).
4. Tumor cell fraction/copy number analysis
The TSO500 assay does not employ an algorithmic process to calculate TCF; rather, it incorporates estimations provided by pathologists into the final report. In contrast, the TE-WGS approach utilizes a computational algorithm that considers CNVs and allelic balance changes. While there exists a correlation between TCF calculations using the TE-WGS method and pathologists’ estimates (r=0.44), a discernible discrepancy in estimation between the two methodologies is evident. This observed incongruity can be attributed to the well-documented tendency of pathologists to provide higher TCF estimations compared to those derived from the WGS-based approach (Fig. 4A) [28].
The copy number of gene amplifications reported by TSO500 was overall concordant with the TE-WGS (r=0.723, p=4.19e-08) (Fig. 4B). Given that tumor purity determined by pathologists is used to normalize the gene copy number of TSO500, we recalculated the copy number using tumor purity more accurately determined by TE-WGS. As a result, the copy numbers between TSO500 and TE-WGS showed a higher correlation (r=0.842, p=7.24e-11) (Fig. 3C, S1 Fig.). All the tier I and II CNVs from TSO500 were also detected by TE-WGS. In addition, TE-WGS identified six deletions that were not reported by TSO500: four CDKN2A, one TP53, and one BAP1 deletion.
5. Gene fusions analysis
TSO500 identified two fusion events that were actionable by the CKB (Jackson Laboratory). These fusion events were associated with an Association for Molecular Pathology/American Society of Clinical Oncology/College of American Pathologists Consensus Recommendation level A or D [29]: CD74::ROS1 in lung adenocarcinoma and TMPRSS2::ERG in prostate cancer. TE-WGS approach reported both fusions from WGS data (Fig. 3D).
6. Genomic instability analysis
TMB: The TSO500 assay estimates TMB by calculating the number of passing eligible variants and dividing it by the size of the targeted coding region covered by the panel. In contrast, the TE-WGS approach employs a germline subtraction method to determine the number of somatic alterations per million base pairs across the entire genome. A TMB value exceeding 10 mutations per megabase (Mb) is classified as ‘high’ (≥ 10 mut/Mb) [30]. TMB scores calculated by TE-WGS exhibited a good correlation with those derived from TSO500 (r=0.89) (Fig. 5A). However, some instances of discordance were observed. Notably, six TSO500-tested tumors classified as TMB-high (12.2%) were categorized as TMB-low (< 10 mutations/Mb) by the TE-WGS approach (Fig. 5B).
MSI: TSO500 relies on stitched reads obtained from tumor samples to determine the MSI status. Unstable microsatellite sites are detected by assessing the shift in the length of a microsatellite site for a tumor sample against a set of normal baseline samples. In contrast, TE-WGS employs a proprietary that incorporates germline sequence subtraction and detects somatic microsatellite changes for MSI estimation. There was a strong correlation between these two methodologies in MSI scoring (r=0.96) (Fig. 5C). The majority of tumors in the cohort were MSI-stable. However, there was one ureter cancer that was identified by TSO500 as MSI-high and confirmed to be MSI-high by TE-WGS. Further, we detected MSH2 (NM_000251.3:c.1043del, p.Gln348ArgfsTer9) somatic mutation, which likely underlies the MSI-high phenotype in the case.
HRD: TSO500 does not include HRD evaluation, and Illumina offers an optional add-on kit for HRD assessment. However, this supplementary kit was not integrated into our study design, thus preventing a direct head-to-head comparison of HRD scoring between the two methodologies. The TE-WGS approach employed a proprietary algorithm dedicated to HRD scoring. Notably, this method effectively discriminated between homologous recombination (HR)–proficient and HR-deficient tumors (Fig. 5D). Four of six HR-deficient tumors harbored germline or somatic mutations in HRD-related genes (BRCA1/2) accompanied by loss of non-mutant allele. While TSO500 did detect BRCA1/2 variants in cases identified by TE-WGS, it did not provide an informative HRD assessment.
7. Mutational signature
WGS is an unbiased approach to profiling mutational signatures [31]. Mutational signature analysis identified tumor samples with HRD, characterized by enrichment of single base substitution (SBS) signature 3 and small insertions and deletions signature 6 (Fig. 6). Mutational signature analysis also identified tumor samples with APOBEC (apolipoprotein B mRNA-editing catalytic polypeptide) mutations, characterized by enrichment of SBS2 and SBS13 signatures (Fig. 6). Identification of the HRD mutational signature suggests that the mutational process is caused by HRD, even if the related genes, such as BRCA1/2, are not found.
Discussion
Cancer remains a formidable challenge in modern medicine despite remarkable advances in treatment modalities such as targeted therapies and immunotherapies. These breakthroughs have ushered in the era of precision medicine, substantially improving survival rates across various cancer types [28]. However, the clinical application of these therapies hinges on the accurate identification of actionable genomic alterations. NGS technologies have demonstrated their clinical utility by enabling the identification of such alterations, facilitating informed decisions regarding approved therapies, clinical trial participation, or the management of rare actionable mutations [32,33].
Nonetheless, a crucial limitation of fixed-panel NGS assays like TSO500 has hindered their ability to keep pace with new discoveries. These panels are inherently constrained by the genomic alterations known at the time of their design. As new biomarkers emerge, these tests quickly become outdated, depriving patients of access to the latest insights that could guide their care. Moreover, the development and validation of updated panels introduce a lag that impedes the timely adoption of emerging biomarkers [11,34,35], necessitating novel approaches.
TE-WGS offers an intriguing solution to this predicament. By combining a deep-read panel with a WGS framework, TE-WGS provides the most comprehensive view of the genomic landscape, unburdened by the constraints of predefined gene panels. This adaptability allows for swift algorithm updates without creating and validating entirely new tests. Additionally, the WGS approach inherently can retrospectively identify emerging biomarkers in previously sequenced patient data, ensuring that patients benefit from the latest discoveries, even after their initial testing. Moreover, the target-enhanced aspect of TE-WGS significantly contributes to overcoming the lower sensitivity associated with critical mutations or genes due to low sequencing depth and potential low TCF in samples. This precision targeting ensures that important genomic regions are thoroughly covered, enhancing the detection sensitivity for key mutations. This feature is particularly valuable in scenarios where conventional sequencing methods may fall short in capturing crucial genetic information, thereby bolstering the reliability of TE-WGS in identifying clinically relevant alterations even in challenging samples.
The clinical utility of targeted genomic panels, encompassing both germline susceptibility and known “actionable” somatic mutations (including TSO500), has been well-documented [35-41]. For some cancer types, TPS is becoming standard practice [39]. This study demonstrated a robust correlation between TSO500 and TE-WGS in detecting SNVs, insertions/deletions (indels), and CNVs. This correlation implies that the clinical utility demonstrated by TSO500 can also be applied to the actionable key biomarker findings reported by the TE-WGS test. Furthermore, TE-WGS extends the actionability seen with TSO500 by reporting the origin (somatic or germline) of actionable biomarkers. Distinguishing between these categories allows for more informed treatment decisions, potentially reducing the risk of exposure to ineffective therapeutic approaches.
In recent years, genomic instability markers have gained attention with the advent of immunotherapy. TMB has emerged as a predictive biomarker for immunogenic neoantigens [6,7]. However, studies have suggested that TMB calculations using tumor-only assays can yield falsely elevated results compared to those determined by germline subtraction methodologies [19]. Furthermore, this discrepancy likely arises from the differential mutation rates between coding and non-coding regions, especially when a small portion of variants are considered. Coding regions, which are more likely to harbor mutations that provide a selective advantage in tumors and are often targeted by sequencing panels, tend to be enriched for mutations compared to non-coding regions [42]. The TE-WGS approach employs a proprietary algorithm informed by germline subtraction and whole genome analysis. While this study showed a good correlation between TSO500 and TE-WGS approaches, instances of discordance were observed, indicating that TE-WGS methodology may potentially reduce exposure to ineffective therapeutic approaches.
Beyond the classical clinical utility of targeted panels, the TE-WGS approach can detect somatic variants arising from specific mutagenic mechanisms or signatures. Each mutagenic source produces a characteristic mutational pattern, and unsupervised learning techniques can decipher signatures associated with distinct etiologies. These signatures, often called “genomic scars”, represent lasting and detectable evidence of environmental or internal genomic damage [13,43]. The most notable and actionable mutational signature currently is the HRD signature. Tumors with mutations in BRCA1/2 are deficient in the homologous recombination repair process, and they exhibit promising responses to PARP inhibitors [44]. The TE-WGS approach allows for bioinformatic integration of known mutational signatures (COSMIC mutational signatures [45]) in a clinical setting, eliminating the need for additional testing for each mutational signature.
Economic studies have shown that NGS panel testing offers cost savings for both Centers for Medicare & Medicaid Services and commercial payers compared to traditional limited gene testing [34,46,47]. A decision-analytic model comparing BRCA1/2 testing to a seven-gene panel for women with a family history of cancer found the seven-gene panel to be more cost-effective, with incremental cost-effectiveness ratios of $23,734 and $42,067 per life-year gained [48]. Although parallel germline and somatic testing strategies enhance variant identification [49], they contribute to inefficient spending. The TE-WGS approach provides a cost-effective solution by consolidating germline, somatic, genomic instability, and mutation signature analysis into a single test, streamlining the diagnostic process and reducing overall costs.
In conclusion, TE-WGS emerges as a promising and potent approach in the pursuit of personalized medicine in oncology. It effectively addresses the need for comprehensive profiling while remaining cost-effective, and adaptable to emerging biomarkers.
XML Download