Prevalence of BRCA1, BRCA2, and PALB2 genomic alterations among 924 Taiwanese breast cancer assays with tumor-only targeted sequencing: extended data analysis from the VGH-TAYLOR study

The full protocol of the VGH-TAYLOR study (Veterans General Hospital Taipei—Yung-Ling foundation sinO-canceR study, ClinicalTrials.gov: NCT04626440), which focuses on the heterogeneity of Taiwanese breast cancer patients and included initial targeted sequencing of 380 and 648 assays, has been described elsewhere [32‐34].

All patients had been evaluated by their clinicians upon diagnosis. The treatment options were determined by physicians according to patients’ characteristics, molecular subtypes, and clinical stages. The concept of share-decision making (SDM) was fully informed and the process of SDM was carried out before the initiation of treatment. Enrolled subjects were subsequently assigned into Group 1 [planned to receive surgery as the first-line treatment and followed by adjuvant therapy, Group 2 [planned to receive neoadjuvant therapy as the first-line treatment and followed by surgery], and Group 3 [diagnosed with de novo and treatment naïve stage IV breast cancer, or stage IV breast cancer with recurrence beyond three years after surgery] (Fig. 1). Three years of enrollment and 4 years of follow-up after enrollment were planned.

Informed consent was obtained from all participants after a thorough explanation by investigators (CCH and LMT). Clinical parameters were assessed by immunohistochemistry (IHC) with estrogen receptor (ER) and progesterone receptor (PR) positivity defined as at least 1% of tumor cells exhibiting nuclear staining. Hormone receptor positivity was defined as either ER-positive or PR-positive. Patients with human epidermal growth factor receptor II (HER2) testing scored as IHC 3+ (positive) or 2+ (equivocal) and with fluorescence in situ hybridization (FISH)-confirmed amplification were considered HER2-positive. All patients received treatment according to the contemporary practice guidelines of the Comprehensive Breast Health Center at Taipei Veterans General Hospital, which were based on the NCCN and St. Gallen guidelines [35, 36].

The details of tumor only targeted sequencing have been described elsewhere [32, 33]. For next-generation sequencing (NGS) library preparation, the Ion Torre Oncomine™ Comprehensive Assay v3 (Thermo Fisher Scientific, Waltham, MA) was used, enabling the detection of 161 cancer-related genes and identification of single nucleotide variants (SNVs), copy number variations (CNVs), gene fusions, and indels.

A total of 879 consecutive breast cancer patients, representing 924 assays, were enrolled in the study. Formalin-fixed paraffin-embedded specimens were assayed with sequencing data analyzed using the Torrent Suite software. The data were further aligned and annotated by the Ion Reporter with the default Oncomine BRCA (5.12) filter applied. Software versions were Torrent Suite (v5.10.0), Ion Reporter (v5.10), Coverage Analysis (v5.10.0.3), SampleID (v5.10.0.1), and VariantCaller (v5.10.0.18). In current study, we focused on BRCA1, BRCA2, and PALB2 mutations.

Variants were further filtered with the Oncomine™ Knowledgebase Reporter (Thermo Fisher Scientific). To further correct spurious findings due to transethnic discrepancies, the online VariED tool was consulted to filter out Taiwan Biobank polymorphisms [37]. The mutational consequences of filtered variants were determined with the ClinVAR database, a freely accessible public archive of reports on the relationships between human variations and phenotypes [38].

We also used the OncoKB™ to assess the clinical implications of variants that were not identified or confirmed by the Oncomine™ Knowledgebase Reporter or the ClinVAR. The OncoKB™ is a knowledge base for precision oncology that offers information on the therapeutic implications of particular genetic alterations in cancer. The database is curated by a group of oncologists, researchers, and bioinformaticians who meticulously assess the available evidence related to each variant, including clinical trial results, FDA approvals, and guidelines from professional organizations. The OncoKB™ also provides annotations for the functional and structural impact of each variant, as well as the level of evidence supporting the annotation. By using the OncoKB™ to evaluate variants that other databases have not identified or confirmed, we can gain a more thorough understanding of the potential clinical implications of these variants [39]. Tumor sequencing variant annotations were functionally annotated using the SNPnexus [40].

To ensure quality control, over 90% of amplicons with > 100× coverage were used as a parameter. Additionally, a minimum coverage of 250 × was deemed acceptable for detecting SNVs and indels with allele fractions of 10% and 20%, respectively [41, 42]. Actionability was determined based on the joint consensus recommendation of the Association for Molecular Pathology, American Society of Clinical Oncology, and College of American Pathologists [43].

When patients' tumor samples tested positive for pathogenic or likely pathogenic variants as identified by ClinVar, reflex germline testing was conducted using whole exome sequencing (WES) through blood sample collection. In some cases, whole genome sequencing (WGS) may have been utilized prior to WES. Both WGS and WES are essential in detecting germline mutations linked to a variety of genetic disorders. WGS offers an exhaustive analysis by sequencing the entire genome, allowing for the identification of mutations across both coding and non-coding regions, as well as the detection of structural variations and copy number variations for a detailed examination. In comparison, WES, while more cost-effective and quicker, concentrates on coding regions or exons, potentially overlooking certain mutations that would be detected by WGS.

To distinguish germline from somatic mutations with tumor-only sequencing, we employed algorithms including the LOH-germline inference calculator (LOHGIC) [44] and the somatic-germline-zygosity (SGZ) method as alternatives for patients not ready for germline testing [45]. We integrated both methods, considering tumor purity, allele frequency, ploidy, and copy number variations for mutation classification. Only tumor purity and allele frequency were considered as our cohort showed no copy number variations for BRCA1/2 and all samples were diploid. We simplified both the LOHGIC and the SGZ by directly comparing observed allele frequencies against those expected in germline and somatic mutations.

For germline mutations, expected in both tumor and normal tissues, the allele frequency should be around 50% in pure normal samples and vary in tumor samples due to loss of heterozygosity. In contrast, somatic mutations, present only in tumor cells, will have an allele frequency dependent on tumor purity, typically near 50% in pure tumor samples but lower in samples with less tumor purity intertwined with adjacent normal tissues.

In current study, we presented the updated results of 924 Thermo Fisher (TMO) OCP v3 assays obtained from 879 breast cancer patients, with 43 patients being assayed twice, and 1 patient being assayed thrice from the VGH-TAYLOR study. The patient distributions of clinical scenarios were as follows: Group 1A (surgery first, n = 578, 65.8%); Group 1B (recurrence within 3 years, n = 22, 2.5%); Group 2 (neoadjuvant therapy, n = 117, 13.3%); Group 3–1 (de novo stage IV, n = 40, 4.6%); and Group 3–2 (recurrence beyond 3 years, n = 67, 7.6%). In addition, there were samples of 55 patients (6.3%) from the retrospective biobank cohort. Figure 2 displays the distributions of IHC results and molecular subtypes.

Of the 924 assays, 281 were positive for mutant BRCA1, BRCA2 and PLAB2 in 130 patients. These mutations impacted 3.1% (27 patients), 8.6% (76 patients), and 5.2% (46 patients), respectively. In total, genetic alterations were noted in 14.8% (130 patients). Details of distribution and patterns of BRCA1, BRCA2, and PALB2 mutations were portrayed in Table 1, Figs. 3, 4, and 5.

The BRCA1 mutation cohort is associated with a higher proportion of advanced stages compared to those without. Additionally, the BRCA2-mutant patients show a higher incidence of family history of ovarian cancer, resulting in a significant difference in the number of mutant patients with a family history of ovarian cancer (Table 1).

In terms of IHC phenotypes, 13 (2.3%) of the BRCA1 mutant breast cancers were HR+/HER2−, 3 (3.3%) were HR+/HER2+, 2 (2.6%) were HR−/HER2+, and 9 (6.9%) were HR−/HER2−. For BRCA2 mutated cases, 51 (9.0%) were HR+/HER2−, 9 (9.9%) were HR+/HER2+, 6 (7.9%) were HR−/HER2+, and 9 (6.9%) were HR-/HER2-. Among the PALB2 mutated patients, 30 (5.3%) were HR+/HER2−, 4 (4.4%) were HR+/HER2+, 3 (3.9%) were HR−/HER2+, and 9 (6.9%) were HR−/HER2− (Table 1).

The study revealed the co-occurrence of BRCA1/2 in 13 breast cancer samples (log2 odds ratio: > 3, p-value < 0.001, and q-value < 0.001). Additionally, the co-occurrence of BRCA1 and PALB2 was found in 8 samples (log2 odds ratio: > 3, p-value < 0.001, and q-value < 0.001), and the co-occurrence of BRCA2 and PALB2 was found in 8 samples (log2 odds ratio: 2.401, p-value < 0.001, and q-value < 0.001). These findings are presented in Table 2.

Among these patients, 5 had both BRCA1/2 mutations, 1 had both BRCA2 and PALB2 mutations, none had both BRCA1 and PALB2 mutations, and 7 had all three mutations. The list of these patients and the variants that they harbored are shown in Table 3. Twenty-four patients had two or more variants. Of those, 10 patients had dual variants, two had triple variants, one had four variants, three had five variants, one had seven variants, one had eight variants, one had nine variants, one had 15 variants, one had 16 variants, one had 19 variants, one had 24 variants, and one had 25 variants. (Table 3).

Among all the variants, p.I887fs was observed 30 times in PALB2-mutant assays. For BRCA2-mutant assays, p.S2186fs, p.V2466A, and p.X159_splice were observed 5 times, while p.T912fs and p.E33* were observed 3 times each. In BRCA1-mutated assays, only p.K654fs was observed three times, while other variants were observed no more than twice. It should be noted that although p.N372H was observed 26 times in BRCA2-mutated assays, it has been confirmed to be a benign variant. The list of those recurrent variants with clinical implications is displayed in Fig. 6.

The study analyzed various genetic variants and identified 176 amino acid (AA) changes, and the characteristics stratified by genes and the clinical implications are listed in Table 2. There were four variant that did not notice any AA change, and three novel splice site variants (BRCA1 c.5256+1G > A, BRCA1 c.5215+1G > A, and BRCA2 c.-38-3CAG > C) were identified, all listed in Table 4. Although these novel variants had no AA change, they were classified as Pathogenic/Likely pathogenic (P/LP) by the ClinVAR database. Notably, 60.2% (106) of the discovered AA changes were not documented in either ClinVAR or the Oncomine™ Knowledge database. Using the OncoKB™ for annotation, 171 (97.2%) AA changes were found to have clinical implications. Half of all missense mutations without clinical implications (4 out of 8) were deemed insignificant, and 23.1% (3 out of 13) of AA changes not recorded in the ClinVAR or the Oncomine™ Knowledge database was not considered clinically relevant. The BRCA2 mutation cohort exhibited the highest proportion of P/LP variants. This cohort also contained the only benign variant identified and had the highest number of AA changes without clinical implications, as per the OncoKB™ (Table 5).

In our study, reflex germline testing was conducted for patients until November 20, 2023. These procedures are summarized in Table 6, which concentrates on WGS and WES analyses for individuals who had positive results from tumor-only sequencing. Specifically, 48 cases, constituting 36.9%, were identified with pathogenic or likely pathogenic variants via tumor-targeted sequencing, as classified by the ClinVar database, and were subsequently recalled for further investigation through WGS or WES.Among 130 cases examined, 7 cases (5.4%) completed WGS uncovering crucial genetic variations. None harbored germline mutations in BRCA1/2 and PALB2. Another 20 cases (15.4%) were reached and 9 had completed WES (6.9%). The study group encountered enormous challenges including loss of follow-up and 14 were deceased.

The combined results of whole genome sequencing (WGS) and whole exome sequencing (WES) have provided in-depth insights into the origins of identified variants. Notably, 4 cases (3.1%) exhibited pathogenic germline mutations. Reflex germline testing results showed that three variants (BRCA1 c.1969_1970del, BRCA1 c.3629_3630del, BRCA2 c.8755-1G > C) were classified as Pathogenic/Likely pathogenic (P/LP) by ClinVar and as likely loss-of-function or likely oncogenic by OncoKB. Meanwhile, one variant (PALB2 c.448C > T) was not listed in ClinVar, but OncoKB annotated it as likely loss-of-function or likely oncogenic. Additionally, there were 2 cases (1.5%) of germline mutations with uncertain significance, and 5 cases (3.8%) with benign germline alterations, which emphasize the genetic intricacy involved in the development of breast cancer. The result for all the variants from reflex germline testing is presented in Additional file 1: Table S1.

Table 7 illustrates the results of LOHGIC and SGZ analyses. Among the 281 samples, 40 were identified as germline mutations and 169 samples (60.1%) were of somatic origin. Borderline cases comprised 26 samples (9.3%). Lastly, there were 46 samples, (16.4%) unclassifiable due to missing data in tumor purity.

Recent research has revealed a significant occurrence of harmful variants in three key genes associated with breast cancer risk, as analyzed through tumor genomic profiling. In a significant study by the Breast Cancer Association Consortium in 2021 involving over 60,000 breast cancer cases and 53,000 controls, sequencing was conducted on 34 potential risk genes. The standout genes were BRCA1, BRCA2, and PALB2. Variants leading to truncating/incomplete proteins in these genes correlated with an increased breast cancer risk, which were statistically significant (all P-values less than 0.0001), underscoring their crucial role in the genetic landscape of breast cancer. These genes are also recognized markers for determining the eligibility for PARP inhibitor therapies, warranting further investigation into their contribution to breast cancer risk [46]. The frequency of mutations in these genes varies by the method of testing and the population studied. Research focusing on Asian patients with BRCA1/2 mutations found a wide prevalence range, with BRCA1 mutations present in 2.3–42% and BRCA2 mutations in 2.3–11.4% of the group [47]. In Taiwan, a study showed a 3.8% prevalence of these mutations in an unselected patient population [30]. Reports on PALB2 mutations in Asian populations are scarce; however, a thorough global review indicated that PALB2 pathogenic variants occur in 0.9% to 3.2% [19].

While germline mutations in breast cancer susceptibility genes have been extensively studied, reports of somatic mutations are rare. However, with the advent of tumor-targeted sequencing, we can explore both germline and somatic mutations simultaneously. In 2022, the Dana-Farber/Harvard Cancer Center conducted a tumor-targeted sequencing study across a broad range of malignancies on 7575 patients, which included 1514 breast cancer patients. If a patient was identified with any P/LP variants of BRCA1, BRCA2, or PALB2 within the tumor, clinical germline testing (CGT) would be performed. The study found that BRCA1 and BRCA2 mutations were present in 2.5% and 3.7% of breast cancer patients, respectively, while PALB2 mutations were present in 0.6% of then. Out of all P/LP variants from tumor-sequencing, 70.5% were confirmed as germline mutations [48].

The study analyzed various genetic variants and identified 176 amino acid (AA) changes. All BRCA1, BRCA2, and PLAB2 mutation information was curated and confirmed by the ClinVAR, the Oncomine™ Knowledgebase Reporter, and the OncoKB™. There were 4 variants that did not have AA changes reported, and 3 of them were novel variants (BRCA1 c.5256+1G > A, BRCA1 c.5215+1G > A, and BRCA2 c.-38-3CAG > C), which were reported the first time in this Taiwanese cohort. The most common BRCA1 mutation was p.K654fs (c.1960_1961insG), which is a frameshift insertion and deleterious mutation (three cases). Although not recorded by the ClinVAR or the ACMG 73 genes the OncoKB™ has confirmed its clinical implication. The most common BRCA2 mutations were p.N372H (c.1114A > C, 26 cases), p.S2186fs (c.6556_6557insA; 5 cases), p.V2466A (c.7397T > C; 5 cases), and p.X159_splice (c.476−2A > G/c or 476−3C > T; 5 cases). The variant p.N372H was initially reported in 2000 and is one of the common non-synonymous polymorphisms [49, 50]. It has been a research focus in the scientific community and has drawn increasing attention [50‐58]. A meta-analysis of 22 studies, involving 22,515 cases and 22,388 controls, found no significant association between the BRCA2 p.N372H polymorphism and breast cancer risk. This suggests that the BRCA2 p.N372H allele may be non-pathogenic. Although p.S2186fs is not annotated by the ClinVAR or listed in the ACMG 73 genes, it is a frameshift insertion and deleterious mutation with clinical implications. This has been confirmed by the OncoKB™. For p.V2466A, the ClinVAR provides conflicting interpretations, with some considering it a benign entity. Regarding p.X159_splice, it is interesting to note that there are two different coding variants being recorded. c.476-3C > T has conflicting implications and has been seen in patients from Central/Eastern Europe. In contrast, c.476-2A > G has been ascertained as pathogenic and has been mentioned in Italian and Chinese population. The most common PALB2 mutation observed was p.I887fs (c.2659_2660delAT), a deleterious frameshift mutation that was observed in 30 patients. Although this mutation is not listed in the ClinVAR or ACMG 73 genes list, its clinical implications have been approved by the OncoKB™.

The study revealed a significant tendency of co-occurrence between BRCA1/2, BRCA1-PALB2, and BRCA2-PALB2 mutations. It should be noted that 17 samples were discarded due to missing values in at least one of the interrogated genes, preventing analysis of mutual exclusivity. Constructing a gene–gene interaction (GGI) network is important for understanding breast carcinogenesis, as single gene or protein alterations are not sufficient to induce cancer. Rather, the interactions with other genes or microenvironment play a key role. A large-scale study on the interaction between genes was conducted on European Non-Small Cell Lung Cancer (NSCLC) risk, using a total of 445,221 participants from various projects [59]. The study found important gene–gene interactions in the 5p15.33 and 6p21.32 regions, which can be used to improve lung cancer screening models.

It was found that BRCA1 interacts with RAD51 to play a role in DNA repair, while BRCA2 co-localizes with RAD51 and BRCA1, indicating a similar function [60]. PALB2, first reported by Xia et al. in 2006 [61], plays a fundamental role in HR. It acts as a bridging molecule that connects the BRCA complex (BRCA1-PALB2-BRCA2-RAD51) and facilitates the function of RAD51, a protein that is vital for strand invasion during HR [18, 19].

A large-scale mutational analysis in 7,325 individuals identified four interactions between mutations in the breast cancer susceptibility genes [62]. These interactions include ATM and CHEK2 with BRCA1 and BRCA2, ATM and BRCA, CHEK2 and BRCA1/BRCA2 combined, and CHEK2 and BRCA1 or BRCA2. The results show a lower risk of breast cancer than that predicted by the multiplicative product of the constituent risks. These findings likely reflect functional relationships between the encoded proteins in DNA repair and have important implications for models of disease predisposition and clinical translation.

Currently, limited studies have addressed the gene–gene interaction among BRCA1, BRCA, and PALB2 in real-world settings of large-scale breast cancer gene analysis. Our results may shed light on exploring the association between breast cancer susceptibility genes and the possibility of creating a genetic panel for predicting and prognosing hereditary breast cancer.

The BRCA1 mutation cohort has a higher proportion of advanced-stage disease compared to others, which may be due to various reasons. One possible explanation is that patients with BRCA1 mutations are more likely to have triple-negative breast cancer [63], which is associated with a higher risk of distant recurrence [64, 65]. Another possible reason is that BRCA1 mutations may be associated with a higher likelihood of developing bilateral breast cancer, which could increase the risk of disease progression [66]. Additionally, patients with BRCA1 mutations are more likely to develop breast cancer at a younger age, when the disease may be more aggressive and therefore more likely to be diagnosed at an advanced stage [67, 68]. Despite having more advanced disease, evidence is mixed as to whether BRCA-associated breast cancer has poorer outcomes. Some studies have shown that carriers of a BRCA1 mutation have worse overall survival [69], while others have found no significant difference in outcomes [70].

Nowadays, tumor-only targeted sequencing has gained increasing attention. One advantage of tumor-only testing is that it can reveal P/LP variants in genes associated with cancer predisposition and potential therapeutics with a higher level of coverage. Identification of mutations through tumor-only sequencing may lead to reflex germline testing, which can identify individuals at an increased risk for cancer and allow for early detection and intervention.

Recent studies have indicated that direct tumor sequencing could be more advantageous than solely testing for inherited mutations. The comprehensive tumor sequencing effort followed by clinical genetic testing at Dana-Farber/Harvard Cancer Center has not only highlighted the commonness of pathogenic/likely pathogenic mutations in genes such as BRCA1, BRCA2, and PALB2, but has also underscored the significance of this targeted sequencing approach [48]. They found that patients with BRCA mutations were more likely to undergo CGT than those without. Interestingly, over half (52.9%) of the tumor-identified P/LP patients did not meet any personal or family history criteria for CGT. Additionally, 32.7% of patients with BRCA1/2 or PALB2 P/LP variants did not have any other clinical indication for germline testing. Nonetheless, 70.5% of P/LP variants identified through CGT were germline origin. These results show the potentiality of tumor-only sequencing in detecting P/LP mutations in cancer predisposition genes across malignancies. Furthermore, they highlighted the necessity of expanding the indications for CGT beyond traditional criteria. A significant proportion of patients with these mutations may not have any personal or family history of cancer, which may have a significant impact on cancer risk assessment, surveillance, and treatment decisions in the future. Before universal germline and tumor sequencing becomes feasibility, tumor-targeted sequencing seems to be a reasonable choice for personalized therapy.

Both germline and somatic alterations can affect treatment decisions and outcomes. For breast cancer patients, the results of the TBCRC-048 and TBB trials suggested further exploration of PARP inhibitors in metastatic or advanced breast cancers with HR-associated mutations beyond BRCA1 and BRCA2 [16, 17]. Identifying additional biomarkers to expand this treatment in somatic BRCA1/2-mutant or HR-related-gene-mutant advanced breast or ovarian cancers could significantly benefit patients who would otherwise receive chemotherapies as the only regimen. These efforts may reveal a patient population that would benefit from targeted therapy, improving patient outcomes and reducing the complications associated with cytotoxic chemotherapy.

The significance and clinical applicability of tumor-targeted sequencing is well acknowledged, yet the unique importance of reflex germline testing cannot be overemphasized. We are continuously expanding cases for such testing. However, the extent of reflex testing conducted to date is still limited, with only 7 cases underwent WGS and 9 with WES (Table 6). This has led us to explore alternative methods to determine whether the reported mutation is germline or somatic origin.

The refined LOHGIC and SGZ methodologies are designed to assess three crucial factors: tumor purity (the proportion of cancer cells in a sample), allele frequency (the incidence of mutations in DNA sequencing reads), and confirmation of a diploid genome (maintaining two copies of each gene). These methods are crucial for differentiating between somatic mutations, which occur in somatic (non-reproductive) cells, and germline mutations, which are inherited. Although these algorithms are useful, they have their limitations; in an analysis of 130 cases, 10 could not be conclusively classified as either germline or somatic due to varying origins of the mutations. Nonetheless, there was a noticeable pattern, with 60% of the mutations being identified as somatic. This observation highlights the imperative for more refined techniques to accurately determine the origins of mutations.

Accurate interpretation of genetic variants is critical in both clinical and research settings. Before reporting detected variants, appropriately trained and certified molecular diagnostic procedures must be carefully carried out in the context of clinical scenarios, including histologic features [43]. However, the classification criteria can vary between submitters, and the evidence for a particular variant may be conflicting, leading to difficulties in an unbiased interpretation. Numerous studies have explored potential indicators for reinterpreting pathogenic variants within specific databases, as well as across distinct platforms. For example, in a recent comparison of the RefSeq and Gencode human gene databases, only 27.5% of transcripts annotated in the Gencode are shared by the RefSeq [71]. Whiffin et al. selected 43 variants from the ClinVAR classified as P/LP, which were not rare enough in at least one of the Exome Aggregation Consortium populations [72]. Their analysis showed that 42 of these variants should be considered variants of uncertain significance (VUS) instead of P/LP. Xiang et al. analyzed common P/LP variants in the ClinVAR database and identified indicators associated with reclassification, indicating missing opportunities due to misinterpretation [73]. The study selected 217 variants in 173 genes for manual interpretation according to guidelines, with 40% of which downgraded to benign or VUS, while 2% identified as more likely risk alleles. Inappropriate classification was associated with low-rank, older annotation, higher allele frequency, and collection through methods other than clinical testing. It is important to note that the reinterpretation of cancer predisposition genes requires a multidisciplinary effort involving clinicians, genetic counselors, bioinformaticians, and researchers [74].

In summary, the reinterpretation of cancer predisposition genes due to annotation inconsistencies among distinct database is an important issue that requires careful consideration and multidisciplinary collaboration. The use of updated annotation and rigid guideline follow-up and the integration of multiple types of genomic data can help improve the accuracy of cancer risk assessment and inform personalized prevention and treatment strategies.

This study analyzing BRCA1, BRCA2, and PALB2 mutations through tumor-targeted sequencing boosts several notable strengths. Firstly, it is the first large-scale analysis of its kind in Taiwan and Asia, involving a substantial number of breast cancer cases. Additionally, the study benefits from the availability of data on family history, molecular subtypes, and early or advanced breast cancer status. Moreover, by utilizing updated annotation databases and guidelines during interpretation and annotation, the study achieves enhanced comprehensiveness and accuracy in its results.

There are several limitations to consider. First, additional germline sequencing has been conducted with compromised recalled rates. Second, the sample size varied among different clinical groups, which may pose a challenge. Third, current study confined to only three genes, potentially narrowing the assessment of the genetic landscape. Lastly, a more in-depth evaluation of clinical outcomes and subgroup analyses, incorporating clinical-pathological factors and treatment regimens, is essential for a deeper understanding of Taiwanese breast cancer and should be conducted in future studies.