A total of 28 female individuals diagnosed with breast cancer participated in this study, and blood samples were obtained from each participant. DNA extraction was conducted using the Genomic DNA Mini Kit® (Geneaid, New Taipei City, Taiwan), following the manufacturer’s instructions. The purity of DNA isolates was assessed by measuring the 260/280 absorbance ratio using a NanoDrop instrument (Thermo Fisher Scientific, Waltham, MA, United States). The quantification of DNA isolates was performed using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA, United States) on a Qubit® 4.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, United States).

Library preparation was performed utilizing the Illumina AmpliSeq™ Cancer Hotspot Panel v2 (Illumina®, United States). The first step involved the amplification of specific areas within the DNA sample. The amplicons were subsequently subjected to partial digestion using the FuPa reagent. The indexes were ligated using the Ligate program on a thermal cycler. In order to purify the libraries, 30 μL of Agencourt AMPure XP beads (Beckman Coulter™, United States) was added to the reaction mixtures.

Amplification techniques were implemented in order to ensure a sufficient quantity of the libraries. The second round of purification was subsequently performed twice to remove high molecular weight DNA and excess primers, using Agencourt AMPure XP beads (Beckman Coulter™, United States). The libraries were diluted to a final loading concentration and subsequently subjected to sequencing utilizing the Illumina MiSeq technology. Sequencing yielded paired-end libraries in FASTQ format, with a read length of 150 base pairs (bp) for both ends. The data sequences have been deposited in the Sequence Read Archive (SRA) database under BioProject accession number PRJNA998562.

Quality control of the FASTQ data was conducted to evaluate the quality of each sample’s raw reads. FastQC software (Andrews, 2010) was used to perform FASTQ quality assessment. The total number of raw bases and Q30 percentage were determined using the q30 Python programs (https://github.com/dayedepps/q30/tree/master). If the quality of sequence reads was poor, quality-improvement steps were taken. Trimmomatic (Bolger et al., 2014) was used to trim the low-quality reads and remove adapters (ILLUMINACLIP: NexteraPE-PE.FA: 2:30:10, LEADING: 3, TRAILING: 3, SLIDINGWINDOW: 4:15, and MINLEN: 35). Read alignment was performed using BWA-MEM (Li and Durbin, 2009), with GRCh38. p13 as the human reference genome. After alignment, the amplicon mean depth, coverage uniformity, and the percentage of on-target rate were calculated using an in-house script containing Mosdepth (Pedersen and Quinlan, 2018), SAMtools (Danecek et al., 2021), and BEDTools (Danecek et al., 2021) software. The command-line scripts used to calculate the Q30 percentage, amplicon mean depth, coverage uniformity, and the percentage of on-target rate are provided in Supplementary Material.

Variant calling analysis was performed to find likely pathogenic and pathogenic variants in all samples. The workflow followed the methods described by Panigoro et al. (2022) and included read alignment using BWA (Li and Durbin, 2009), SAM-to-BAM conversion using SAMTools (Li et al., 2009), variant calling using GATK (McKenna et al., 2010), and variant annotation using SnpEff and SnpSift (Cingolani et al., 2012). Germline variant classification was conducted using VarSome (Kopanos et al., 2019) (https://varsome.com/), which applies the ACMG classification guidelines. A total score is computed by summing the points from pathogenic rules and subtracting the points from benign rules. The total score is then compared with predefined thresholds to determine the final verdict: pathogenic if greater than or equal to 10, likely pathogenic if between 6 and 9 inclusive, uncertain significance if between 0 and 5, likely benign if between −6 and −1, and benign if less than or equal to −7. The command-line scripts used for variant calling analysis are available on GitHub: https://github.com/fikaparamita04/variant-calling. The most frequently observed pathogenic variants were visualized using MutationMapper (Vohra and Biggin, 2013) (https://www.cbioportal.org/mutation_mapper).

We successfully collected blood samples from 28 patients diagnosed with breast cancer at Cipto Mangunkusumo National Hospital, Jakarta. The patients ranged in age from 40 to 71 years (Table 1). The patients were categorized into four subtypes, namely, Luminal A, Luminal B, HER2-positive and triple-negative breast cancer (TNBC), with the total number of patients being 8, 9, 7, and 4, respectively. Among the subjects, four patients were diagnosed at stage IV.

Raw FASTQ data were quality-checked to ensure high sequencing quality. Given that this was a targeted sequencing study, we evaluated the Q30 percentage, average amplicon depth, coverage uniformity, and target level percentage (Table 2). The Q30 result of 97.71% ± 0.44 indicates high-quality sequencing. The average amplicon depth was also strong, with a score of 1,076 ± 256.36, although variability between samples was relatively high. Coverage uniformity, which measures how evenly sequencing reads are distributed across the genome or a specific region of interest, was excellent across all samples. All samples showed a coverage uniformity score of 1, or close to 1 (0.9901 ± 0.01), indicating that all target bases were covered to the same extent, without regions of significantly higher or lower read depth. The target-level percentage, which reflects the proportion of bases within the targeted regions that were successfully sequenced, was also high at 95.57% ± 0.57. The raw data files in FASTQ format have been archived in the BioProject database under accession number PRJNA998562. These data may serve as a potentially valuable resource for screening gene mutation markers in breast cancer and could aid in predicting treatment efficacy related to specific mutations.

We found the highest germline frameshift mutation in the FBXW7 gene (4:g.152324246del), with a frequency of approximately 35.7%. This variant was predicted as likely pathogenic by VarSome (prediction score = 9), due to the loss of protein functions (Figure 1). Interestingly, this mutation was found in all Luminal B and one HER2-positive patient. FBXW7 is a tumor suppressor that modulates the degradation of oncogenic substrates, including c-Jun, c-Myc, the Notch1 intracellular domain (ICD), and cyclin E, by acting as the substrate recognition protein within the Skp1–Cullin–F-box (SCF) ubiquitin ligase complex. Deletion of chromosome 4q3, which encompasses FBXW7, occurs in approximately 30% of primary breast tumors (Meyer et al., 2020).

In line with previous studies, the deletion mutation in the FBXW7 gene closely resembles the human breast cancer luminal B subtype, characterized by ERα+, PR-, and elevated Ki67 staining (Meyer et al., 2020). Furthermore, Luminal B tumors exhibit the lowest FBXW7 mRNA expression among breast cancer subtypes. Lower FBXW7 expression is associated with a high Ki-67 labeling index and positive cyclin E protein expression, both indicators of proliferation. Breast cancer patients with the greatest FBXW7 gene expression have a longer disease-free survival rate (Yeh et al., 2018). The process by which FBXW7 regulates breast cancer growth, cell cycle, and metastasis involves many signaling pathways and gene interactions. For example, FBXW7-deficient breast tumors inhibit the NF-κB signaling pathway, which normally involves E3 ubiquitin ligase binding and degradation. This inhibition results in enhanced NF-κB DNA-binding activity, promoting tumor development and metastasis (Chen et al., 2023).

As shown in Figure 1, the mutation found in the FBXW7 gene is also reported in the OncoKB database (Chakravarty et al., 2017). According to the database, FBXW7 N598Mfs*30 is a truncating mutation in a tumor suppressor gene and is, therefore, considered likely oncogenic. There is promising scientific and anecdotal clinical evidence supporting the use of lunresertib and camonsertib in patients with FBXW7-mutated solid tumors. Lunresertib is an orally available, small-molecule PKMYT1 inhibitor, while camonsertib is an orally available, small-molecule ATR inhibitor. In the Phase I MYTHIC trial of lunresertib plus camonsertib in patients with advanced tumors harboring CCNE1 amplifications, FBXW7 deleterious mutations, or PPP2R1A deleterious mutations, the lunresertib + camonsertib cohort (n = 59 [n = 17 endometrial; n = 13 colorectal; n = 11 ovarian; n = 3, breast; n = 3, lung; n = 12, other]) showed an overall response rate of 23.6% among all evaluable patients across tumor types (n = 55) (Yap et al., 2023). However, future studies with larger patient populations and integration of multi-omics approaches are needed for precise subtyping and personalized therapy. We hope that our small contribution can help advance precision therapy for breast cancer, particularly in Indonesia.