A clinical sample was obtained from a patient 48-year-old female who initially presented with a parasellar mass, clinically resembling an extra-axial tumour. An initial MRI performed in December 2022 revealed a 3.6x4.3x3.1 cm tumour in left parasellar region, raising suspicion for meningioma or craniopharyngioma (Fig. 1A). The patient initially declined surgical intervention and was instead treated with Gamma Knife stereotactic radiosurgery, receiving a dose of 14 Gy in December 2022. Subsequent MRI follow-ups were conducted every six months. The tumour remained stable in size for up to one year post-radiation, as shown in MRI scans from June 2023 and December 2023 (Fig. 1B and C, respectively). After ~1 year and 3 months post-radiation treatment, a brain MRI performed in March 2024 revealed a progressive tumour at the same location extending to the left cerebellopontine angle and abutting the brainstem (Fig. 1D). Surgery was offered and the patient agreed. A subtotal resection was performed in March 2024; however the surgeon was unable to achieve gross total resection due to the tumour encasing the cavernous sinus. At the follow-up 6 months after the surgery, the patient did not exhibit any significant symptoms, and an MRI scan performed in October 2024, revealed stable disease (Fig. 1E). An MRI was performed again 6 months later in April 2025 and revealed a slight reduction of tumour size (Fig. 1F). Since there was no progression of disease, no further intervention was undertaken.

A tissue sample was obtained during craniotomy performed in in March 2024 from the left retroclival region. Histopathological analysis and further immunohistochemical assessment performed in April 2024 were both inconclusive; therefore, further molecular examination was performed. Ethical clearance was obtained from the Ethics Committee of the Faculty of Medicine, University of Indonesia (Jakarta, Indonesia) in accordance with the Declaration of Helsinki, for the collection and molecular testing of biological specimens from rare and recurrent tumours at any site. Written informed consent was obtained from the patient prior to molecular testing, ensuring full disclosure of the purpose, procedures and potential use of biological specimens for research in the present study. The specimen was retrieved from the Department of Pathology of Cipto Mangunkusumo National General Hospital (Jakarta, Indonesia) in the form of formalin-fixed paraffin-embedded (FFPE) tissue. DNA extraction was performed on the FFPE tissue sample, followed by NGS sequencing using a third generation sequencing platform, and the methylation profile was analysed. The DNA extraction and NGS were performed in August 2024. The molecular information was then matched with available clinical, imaging, histopathology and immunohistochemcial results.

The tumour tissue sample was initially fixed in 10% neutral buffered formalin for 24 at 4˚C to preserve cellular structure and protein antigenicity, followed by dehydration and embedding in paraffin wax. Once embedded, the tissue was sectioned into thin slices with a thickness of 4-5 µm using a microtome, and was then mounted onto positively charged glass slides to ensure optimal adherence. The slides were baked at 60˚C to solidify tissue attachment before further processing. The next step was deparaffinization and rehydration. Tissue sections were deparaffinized by immersing them in xylene, followed by rehydration through a graded series of ethanol solutions of decreasing concentrations (100, 95 and 70%) and finally in distilled water. To expose the antigenic sites, antigen retrieval was performed using a heat-induced epitope retrieval (HIER) technique in an EDTA buffer (pH 9.0). This step resulted in unmasking the target epitopes and enhancing antibody binding.

Following antigen retrieval, non-specific binding was minimized by incubating the slides with a blocking solution, 10% bovine serum albumin (cat. no. 37520; Thermo Fisher Scientific, Inc.) for 30 min at room temperature. After blocking, primary antibodies (Table I) specific to each target protein marker were applied at their optimized dilutions following each antibody kit manufacturer guideline. The slides were then incubated in a humid chamber for 2 h at room temperature. This step enabled the primary antibodies to bind specifically to their corresponding antigens in the tissue sections. The detection of bound primary antibodies was carried out using an appropriate secondary antibody conjugated to a biotin enzyme, horseradish peroxidase (HRP) (mouse anti-human IgG4 Fc, secondary antibody, HRP; 1:1,000; cat. no. A-10654; Thermo Fisher Scientific, Inc.) for 1 h at room temperature. A chromogenic substrate, 3,3'-diaminobenzidine (DAB), was used to develop a brown colour, indicating the presence of the target antigen. Counterstaining with hematoxylin was performed for 5 min at room temperature to provide contrast by staining cell nuclei blue, aiding in the morphological assessment of the tissue.

After staining, the slides underwent a series of dehydration steps using graded alcohols, followed by clearing in xylene, and were finally mounted with a permanent mounting medium. The prepared slides were then examined under a light microscope by a senior neuropathologist to evaluate the expression patterns of the specific markers. The staining intensity and proportion of positive cells were recorded for each marker. The immunohistochemical markers examined were Vimentin, AE1/3, EMA, p63, S100, PR, CD34, Ki-67, p40, CK7, CK8, HMWCK, Cam5.2, CK20, ER, STAT6, TTF-1, BRAF V600E, β-catenin, TLE-1, BCL2, CD99, and CD56.

The DNA extraction of the FFPE tissue was carried out using the QIAamp DNA FFPE Tissue Kit (cat. no. 56404; QIAGEN) which involved several key steps. Initially, the FFPE tissue sample was deparaffinized using a xylene and ethanol wash to remove any residual paraffin. Then the tissue was prepared for next process, lysis. This lysis step involved breakdown of the cross-linked proteins in the FFPE samples. The sample underwent incubation with a lysis buffer and proteinase K (both included in the aforementioned kit) to digest the tissue and release the DNA into the solution.

Once lysis was completed, the sample was treated with a heat incubation step up to 90˚C to reverse formalin-induced crosslinking, which helped to ensure optimal DNA recovery and purity. The lysate was then mixed with ethanol and passed through a QIAamp MinElute column, which selectively bonded the DNA. Following several washes to remove contaminants, the purified DNA was eluted in a low-salt buffer, ready for downstream sequencing applications. The authors adhered strictly to the manufacturer-recommended protocol for extracting FFPE tissue. The complete and detailed protocol was available publicly in the corresponding kit documentation online. The extracted DNA was measured for purity using Nanodrop Spectrophotometer and the quantity was measured using Qubit Fluorometers (both from Thermo Fisher Scientific, Inc.).

The extracted DNA was sequenced using a third generation sequencer from (Oxford Nanopore Technologies plc). A ligation-based protocol was used to prepare the library DNA [Ligation sequencing DNA V14 (SQK-LSK114); Oxford Nanopore Technologies plc]. There was no PCR step involved to preserve the native methylation information within the DNA. The type of sequencing was direct DNA sequencing with modified bases for methylation detection. The DNA was sequenced using PromethION flow cell (cat. no. FLO-PRO114M; Oxford Nanopore Technologies plc) with chemistry version R 10.4.1. The library loading concentration was 20 fmol and the software used for analysis (basecalling and alignment) was MinKNOW Version 25.02 (Oxford Nanopore Technologies plc). The detailed library preparation and sequencing steps were carried out according to the publicly available Nanopore protocol on the Nanopore Community webpage. The present study aimed to obtain low coverage whole genome sequencing, and thus the sequencing was carried out for 3 h until the raw data in POD5 format reached ~50 GB. The sequencing was then interrupted and stopped. The overview process for molecular testing, from IHC and methylation testing to correlation with clinical and imaging data, is illustrated in Fig. 2.

After sequencing was completed, post-run basecalling was performed using the dna_r10.4.1_e8.2_400bps_sup@v5.0.0 model, which included 5mCG and 5hmCG base modification calling with Dorado (18). The methylation analysis was then carried out using NanoDx pipeline, which is publicly available (12,19). Briefly, the basecalled sequencing reads were aligned to the hg19 human reference genome using Minimap2(20). Methylation information was then extracted using Modkit version 0.4.0 (https://github.com/nanoporetech/modkit), which provides single-base resolution of methylation data. The methylation matrix was then constructed representing all CpG sites. Furthermore, dimensionality reduction using t-SNE was applied to visualise and analyse the high-dimensional methylation data. The methylation data was classified using the public Heidelberg brain tumour classifier v11b4 reference set (21).

The histopathology results revealed a tumour mass that under microscopic examination, was arranged in irregular short lines, forming sheets, lobules and a patternless architecture (Fig. 3A). The tumour cells exhibited round to oval nuclei with spindle shapes, mild to moderate pleomorphism, slightly coarse chromatin, and eosinophilic cytoplasm. The mitotic count was 6 per 10 low-power fields. Blood vessels appeared partially congested. There was dense infiltration of macrophages or foamy cells around the tumour. This histopathological feature was inconclusive, with the histology suggesting the possibility of meningioma, solitary fibrous tumour, pituitary adenoma or carcinoma.

IHC was performed and the results obtained from all the tested markers are summarized in Table II, Fig. 3B-H, and Fig. S1. The Ki-67 index of up to 20% was indicative of moderately high proliferative activity, suggesting a potentially aggressive tumour. Positive BRAF V600E expression suggested the presence of this mutation in the tumour, which was also indicative of a more aggressive tumour and poorer prognosis. The findings of IHC indicated partial epithelial and mesenchymal marker expression (Vimentin, AE1/3 and EMA) and BRAF V600E positivity, along with negative markers for neural and neuroendocrine differentiation, suggesting that the tumour may be a type of mixed epithelial-mesenchymal tumour or a variant of glioma or meningioma with complex differentiation. The neuropathologist concluded that this was more likely a craniopharyngioma. The ambiguous morphological findings on the histopathology and atypical expression were likely due to prior high-dose stereotactic radiotherapy to the tumour. Based on the histopathology and immunohistochemical results alone, a definitive classification could not be made; however, the most likely diagnosis was craniopharyngioma.

Further molecular analysis was performed using NGS to conduct direct sequencing in order to detect random whole genome methylation from the tumour DNA. The methylation classification analysis of the CNS tumour sample identified the tumour as a meningioma at both the methylation class and methylation class family levels. The confidence score for this classification was 0.791, which was above the commonly used threshold of 0.2, indicating a reliable prediction. The Heidelberg classifier, validated in large cohorts, demonstrates a sensitivity of 92-95% and specificity of 90-93% for meningioma classification in untreated cases (22-24), although its performance in irradiated tumours remains less well studied. In this context, the 0.791 score suggests reliable discrimination despite post-treatment complexity. This results suggested that the epigenetic profile of the tumour was most consistent with meningioma when compared to the reference profiles of CNS tumours. Additionally, the classification was supported by the absence of strong signals for other CNS tumour entities among the top five categories, which included high-grade neuroepithelial tumours, atypical teratoid/rhabdoid tumours, and various gliomas (Fig. 4A).

The copy number variation (CNV) profile, shown in Fig. 4B, was generated based on low-coverage whole genome sequencing data. Due to the low coverage, the resolution of the CNV analysis may be limited, and fine-scale alterations could be missed. Nonetheless, it provided an overview of the broader chromosomal changes present in the tumour specimen. Overall, the CNV plot revealed regions with potential copy number gains (positive log2 values represented by red bars) and losses (negative log2 values represented by blue bars) across multiple chromosomes. The copy number profile showed relative stability across most chromosomes, with the majority of regions falling within the-0.5 to 0.5 range on the y-axis, indicating no major large-scale chromosomal gains or losses.

Chromosome 5 and 6 exhibited large segments of copy number loss, indicating possible deletions. A slight gain was visible on the q arm of chromosome 1, extending approximately from the middle to the end of the chromosome. The 1q gain observed in the copy number profile was consistent with the classification of a meningioma, as gain of 1q is a known recurrent alteration in certain meningioma subtypes (25). Chromosome 22 had a noticeable deletion, which was consistent with known chromosomal patterns in meningiomas (26,27). Despite the lower resolution of this analysis, these broad CNV changes provide valuable insights into the overall genomic landscape of the tumour and could serve as a preliminary indicator for further validation using higher-coverage sequencing methods.

The dimensionality reduction plot further corroborated this classification by visually clustering the sample close to the meningioma reference group, distinct from other tumour classes such as glioblastomas or embryonal tumours (Fig. 4C). This visualization tool, while only qualitative, reinforces the classification outcome by showing that the methylation pattern of the sample overlaps predominantly with meningioma profiles. Given the relatively high number of CpG sites used in the classification (11,653 sites), the prediction was considered robust, as it comprehensively covered relevant loci for distinguishing CNS tumour types. A final diagnosis was obtained following multidisciplinary assessment of all available data, and considered this entity to be a meningioma. Compared with IHC, which required 5-10 days and yielded inconclusive results, methylation profiling provided a definitive diagnosis in 4-7 days (from DNA extraction to classification), with an estimated accuracy of >90% based on classifier benchmarks (21,23,24). This quantitative advantage highlights the superiority of methylation profiling in both speed and diagnostic precision, particularly in complex cases.