Cerebellar Mutism Syndrome and Dentato-Thalamo-Cortical Tract Disruption in Diffusion Tractography Following Surgery for Medulloblastoma.

Background Cerebellar mutism syndrome (CMS), a complication following medulloblastoma surgery, has been linked to dentato-thalamo-cortical tract (DTCT) injury; the association of the degree of DTCT injury with severity of CMS-related symptoms has not been investigated. Purpose To investigate the association between severity of CMS-related symptoms and degree and patterns of DTCT injury with use of diffusion tensor imaging (DTI), and if laterality of injury influences neurologic symptoms. Materials and Methods This retrospective case-control study used prospectively collected clinical and DTI data on patients with medulloblastoma enrolled in a clinical trial (between July 2016 and February 2020) and healthy controls (between April and November 2017), matched with the age range of the participants with medulloblastoma. CMS was divided into types 1 (CMS1) and 2 (CMS2). Multivariable logistic regression was used to investigate the relationship between CMS likelihood and DTCT injury. Results Overall, 82 participants with medulloblastoma (mean age, 11.0 years ± 5.2 [SD]; 53 male) and 35 healthy controls (mean age, 18.0 years ± 3.06; 18 female) were included. In participants with medulloblastoma, DTCT was absent bilaterally (AB), absent on the right side (AR), absent on the left side (AL), or present bilaterally (PB), while it was PB in all healthy controls. Odds of having CMS were associated with higher degree of DTCT damage (AB, odds ratio = 272.7 [95% CI: 269.68, 275.75; P < .001]; AR, odds ratio = 14.40 [95% CI: 2.84, 101.48; P < .001]; and AL, odds ratio = 8.55 [95% CI: 1.15, 74.14; P < .001). Left (coefficient = -0.07, χ2 = 12.4, P < .001) and right (coefficient = -0.15, χ2 = 33.82, P < .001) DTCT volumes were negatively associated with the odds of CMS. More participants with medulloblastoma with AB showed CMS1; unilateral DTCT absence prevailed in CMS2. Lower DTCT volumes correlated with more severe ataxia. Unilateral DTCT injury caused ipsilateral dysmetria; AB caused symmetric dysmetria. PB indicated better neurologic outcome. Conclusion The severity of CMS-associated mutism, ataxia, and dysmetria was associated with DTCT damage severity. DTCT damage patterns differed between CMS1 and CMS2. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Dorigatti Soldatelli and Ertl-Wagner in this issue.

Ganglioglioma with anaplastic/high-grade transformation: Histopathologic, molecular, and epigenetic characterization of 3 cases.

Ganglioglioma (GG) with anaplasia (anaplastic ganglioglioma) is a rare and controversial diagnosis. When present, anaplasia involves the glial component of the tumor, either at presentation or at recurrence. To date, most published cases lack molecular characterization. We describe the histologic and molecular features of 3 patients presenting with BRAF p. V600E-mutant GG (CNS WHO grade 1) with high-grade glial transformation at recurrence. The tumors occurred in pediatric patients (age 9-16 years) with time to recurrence from 20 months to 7 years. At presentation, each tumor was low-grade, with a BRAFV600E-positive ganglion cell component and a glial component resembling pleomorphic xanthoastrocytoma (PXA) or fibrillary astrocytoma. At recurrence, tumors resembled anaplastic PXA or high-grade astrocytomas without neuronal differentiation. CDKN2A homozygous deletion (HD) was absent in all primary tumors. At recurrence, 2 cases acquired CDKN2A HD; the third case showed loss of p16 and MTAP immunoexpression, but no CDKN2A/B HD or mutation was identified. By DNA methylation profiling, all primary and recurrent tumors either grouped or definitely matched to different methylation classes. Our findings indicate that malignant progression of the glial component can occur in GG and suggest that CDKN2A/B inactivation plays a significant role in this process.

Group 3 medulloblastoma transcriptional networks collapse under domain specific EP300/CBP inhibition.

Chemical discovery efforts commonly target individual protein domains. Many proteins, including the EP300/CBP histone acetyltransferases (HATs), contain several targetable domains. EP300/CBP are critical gene-regulatory targets in cancer, with existing high potency inhibitors of either the catalytic HAT domain or protein-binding bromodomain (BRD). A domain-specific inhibitory approach to multidomain-containing proteins may identify exceptional-responding tumor types, thereby expanding a therapeutic index. Here, we discover that targeting EP300/CBP using the domain-specific inhibitors, A485 (HAT) or CCS1477 (BRD) have different effects in select tumor types. Group 3 medulloblastoma (G3MB) cells are especially sensitive to BRD, compared with HAT inhibition. Structurally, these effects are mediated by the difluorophenyl group in the catalytic core of CCS1477. Mechanistically, bromodomain inhibition causes rapid disruption of genetic dependency networks that are required for G3MB growth. These studies provide a domain-specific structural foundation for drug discovery efforts targeting EP300/CBP and identify a selective role for the EP300/CBP bromodomain in maintaining genetic dependency networks in G3MB.

NAB2::STAT6 fusions and genome-wide DNA methylation profiling: Predictors of patient outcomes in meningeal solitary fibrous tumors.

Meningeal solitary fibrous tumors (SFT) are rare and have a high frequency of local recurrence and distant metastasis. In a cohort of 126 patients (57 female, 69 male; mean age at surgery 53.0 years) with pathologically confirmed meningeal SFTs with extended clinical follow-up (median 9.9 years; range 15 days-43 years), we performed extensive molecular characterization including genome-wide DNA methylation profiling (n = 80) and targeted TERT promoter mutation testing (n = 98). Associations were examined with NAB2::STAT6 fusion status (n = 101 cases; 51 = ex5-7::ex16-17, 26 = ex4::ex2-3; 12 = ex2-3::exANY/other and 12 = no fusion) and placed in the context of 2021 Central Nervous System (CNS) WHO grade. NAB2::STAT6 fusion breakpoints (fusion type) were significantly associated with metastasis-free survival (MFS) (p = 0.03) and, on multivariate analysis, disease-specific survival (DSS) when adjusting for CNS WHO grade (p = 0.03). DNA methylation profiling revealed three distinct clusters: Cluster 1 (n = 38), Cluster 2 (n = 22), and Cluster 3 (n = 20). Methylation clusters were significantly associated with fusion type (p < 0.001), with Cluster 2 harboring ex4::ex2-3 fusion in 16 (of 20; 80.0%), nearly all TERT promoter mutations (7 of 8; 87.5%), and predominantly an "SFT" histologic phenotype (15 of 22; 68.2%). Clusters 1 and 3 were less distinct, both dominated by tumors having ex5-7::ex16-17 fusion (respectively, 25 of 33; 75.8%, and 12 of 18; 66.7%) and with variable histological phenotypes. Methylation clusters were significantly associated with MFS (p = 0.027), but not overall survival (OS). In summary, NAB2::STAT6 fusion type was significantly associated with MFS and DSS, suggesting that tumors with an ex5::ex16-17 fusion may have inferior patient outcomes. Methylation clusters were significantly associated with fusion type, TERT promoter mutation status, histologic phenotype, and MFS.

A spatial cell atlas of neuroblastoma reveals developmental, epigenetic and spatial axis of tumor heterogeneity.

Neuroblastoma is a pediatric cancer arising from the developing sympathoadrenal lineage with complex inter- and intra-tumoral heterogeneity. To chart this complexity, we generated a comprehensive cell atlas of 55 neuroblastoma patient tumors, collected from two pediatric cancer institutions, spanning a range of clinical, genetic, and histologic features. Our atlas combines single-cell/nucleus RNA-seq (sc/scRNA-seq), bulk RNA-seq, whole exome sequencing, DNA methylation profiling, spatial transcriptomics, and two spatial proteomic methods. Sc/snRNA-seq revealed three malignant cell states with features of sympathoadrenal lineage development. All of the neuroblastomas had malignant cells that resembled sympathoblasts and the more differentiated adrenergic cells. A subset of tumors had malignant cells in a mesenchymal cell state with molecular features of Schwann cell precursors. DNA methylation profiles defined four groupings of patients, which differ in the degree of malignant cell heterogeneity and clinical outcomes. Using spatial proteomics, we found that neuroblastomas are spatially compartmentalized, with malignant tumor cells sequestered away from immune cells. Finally, we identify spatially restricted signaling patterns in immune cells from spatial transcriptomics. To facilitate the visualization and analysis of our atlas as a resource for further research in neuroblastoma, single cell, and spatial-omics, all data are shared through the Human Tumor Atlas Network Data Commons at www.humantumoratlas.org.

Prognostic value of DNA methylation subclassification, aneuploidy, and CDKN2A/B homozygous deletion in predicting clinical outcome of IDH mutant astrocytomas.

INTRODUCTION: IDH mutant astrocytoma grading, until recently, has been entirely based on morphology. The 5th edition of the Central Nervous System WHO introduces CDKN2A/B homozygous deletion as a biomarker of grade 4. We sought to investigate the prognostic impact of DNA methylation-derived molecular biomarkers for IDH mutant astrocytoma. METHODS: We analyzed 98 IDH mutant astrocytomas diagnosed at NYU Langone Health between 2014 and 2022. We reviewed DNA methylation subclass, CDKN2A/B homozygous deletion, and ploidy and correlated molecular biomarkers with histological grade, progression free (PFS), and overall (OS) survival. Findings were confirmed using two independent validation cohorts. RESULTS: There was no significant difference in OS or PFS when stratified by histologic WHO grade alone, copy number complexity, or extent of resection. OS was significantly different when patients were stratified either by CDKN2A/B homozygous deletion or by DNA methylation subclass (p-value=0.0286 and 0.0016, respectively). None of the molecular biomarkers were associated with significantly better progression free survival (PFS), although DNA methylation classification showed a trend (p-value= 0.0534). CONCLUSIONS: The current WHO recognized grading criteria for IDH mutant astrocytomas show limited prognostic value. Stratification based on DNA methylation shows superior prognostic value for OS.

High-grade glioma in infants and young children is histologically, molecularly, and clinically diverse: Results from the SJYC07 trial and institutional experience.

BACKGROUND: High-grade gliomas (HGG) in young children pose a challenge due to favorable but unpredictable outcomes. While retrospective studies broadened our understanding of tumor biology, prospective data is lacking. METHODS: A cohort of children with histologically diagnosed HGG from the SJYC07 trial was augmented with nonprotocol patients with HGG treated at St. Jude Children's Research Hospital from November 2007 to December 2020. DNA methylome profiling and whole genome, whole exome, and RNA sequencing were performed. These data were integrated with histopathology to yield an integrated diagnosis. Clinical characteristics and preoperative imaging were analyzed. RESULTS: Fifty-six children (0.0-4.4 years) were identified. Integrated analysis split the cohort into four categories: infant-type hemispheric glioma (IHG), HGG, low-grade glioma (LGG), and other-central nervous system (CNS) tumors. IHG was the most prevalent (n = 22), occurred in the youngest patients (median age = 0.4 years), and commonly harbored receptor tyrosine kinase gene fusions (7 ALK, 2 ROS1, 3 NTRK1/2/3, 4 MET). The 5-year event-free (EFS) and overall survival (OS) for IHG was 53.13% (95%CI: 35.52-79.47) and 90.91% (95%CI: 79.66-100.00) vs. 0.0% and 16.67% (95%CI: 2.78-99.74%) for HGG (p = 0.0043, p = 0.00013). EFS and OS were not different between IHG and LGG (p = 0.95, p = 0.43). Imaging review showed IHGs are associated with circumscribed margins (p = 0.0047), hemispheric location (p = 0.0010), and intratumoral hemorrhage (p = 0.0149). CONCLUSIONS: HGG in young children is heterogeneous and best defined by integrating histopathological and molecular features. Patients with IHG have relatively good outcomes, yet they endure significant deficits, making them good candidates for therapy de-escalation and trials of molecular targeted therapy.

Mapping pediatric brain tumors to their origins in the developing cerebellum.

BACKGROUND: Distinguishing the cellular origins of childhood brain tumors is key for understanding tumor initiation and identifying lineage-restricted, tumor-specific therapeutic targets. Previous strategies to map the cell-of-origin typically involved comparing human tumors to murine embryonal tissues, which is potentially limited due to species-specific differences. The aim of this study was to unravel the cellular origins of the 3 most common pediatric brain tumors, ependymoma, pilocytic astrocytoma, and medulloblastoma, using a developing human cerebellar atlas. METHODS: We used a single-nucleus atlas of the normal developing human cerebellum consisting of 176 645 cells as a reference for an in-depth comparison to 4416 bulk and single-cell transcriptome tumor datasets, using gene set variation analysis, correlation, and single-cell matching techniques. RESULTS: We find that the astroglial cerebellar lineage is potentially the origin for posterior fossa ependymomas. We propose that infratentorial pilocytic astrocytomas originate from the oligodendrocyte lineage and MHC II genes are specifically enriched in these tumors. We confirm that SHH and Group 3/4 medulloblastomas originate from the granule cell and unipolar brush cell lineages. Radiation-induced gliomas stem from cerebellar glial lineages and demonstrate distinct origins from the primary medulloblastoma. We identify tumor genes that are expressed in the cerebellar lineage of origin, and genes that are tumor specific; both gene sets represent promising therapeutic targets for future study. CONCLUSION: Based on our results, individual cells within a tumor may resemble different cell types along a restricted developmental lineage. Therefore, we suggest that tumors can arise from multiple cellular states along the cerebellar "lineage of origin."

Predicting methylation class from diffusely infiltrating adult gliomas using multimodality MRI data.

Background: Radiogenomic studies of adult-type diffuse gliomas have used magnetic resonance imaging (MRI) data to infer tumor attributes, including abnormalities such as IDH-mutation status and 1p19q deletion. This approach is effective but does not generalize to tumor types that lack highly recurrent alterations. Tumors have intrinsic DNA methylation patterns and can be grouped into stable methylation classes even when lacking recurrent mutations or copy number changes. The purpose of this study was to prove the principle that a tumor's DNA-methylation class could be used as a predictive feature for radiogenomic modeling. Methods: Using a custom DNA methylation-based classification model, molecular classes were assigned to diffuse gliomas in The Cancer Genome Atlas (TCGA) dataset. We then constructed and validated machine learning models to predict a tumor's methylation family or subclass from matched multisequence MRI data using either extracted radiomic features or directly from MRI images. Results: For models using extracted radiomic features, we demonstrated top accuracies above 90% for predicting IDH-glioma and GBM-IDHwt methylation families, IDH-mutant tumor methylation subclasses, or GBM-IDHwt molecular subclasses. Classification models utilizing MRI images directly demonstrated average accuracies of 80.6% for predicting methylation families, compared to 87.2% and 89.0% for differentiating IDH-mutated astrocytomas from oligodendrogliomas and glioblastoma molecular subclasses, respectively. Conclusions: These findings demonstrate that MRI-based machine learning models can effectively predict the methylation class of brain tumors. Given appropriate datasets, this approach could generalize to most brain tumor types, expanding the number and types of tumors that could be used to develop radiomic or radiogenomic models.

Diffuse leptomeningeal glioneuronal tumor in a child masquerading as an intramedullary spinal pilocytic astrocytoma.

Diffuse leptomeningeal glioneuronal tumor (DLGNT) occurs predominantly in children and is typically characterized by diffuse leptomeningeal lesions throughout the neuroaxis with focal segments of parenchymal involvement. Recent reports have identified cases without diffuse leptomeningeal involvement that retain classic glioneuronal features on histology. In this report, we present a case of a 4-year-old boy with a large cystic-solid intramedullary spinal cord lesion that on surgical biopsy revealed a biphasic astrocytic tumor with sparsely distributed eosinophilic granular bodies and Rosenthal fibers. Next-generation sequencing revealed a KIAA1549-BRAF fusion, 1p/19q codeletion, and lack of an IDH1 mutation. Methylation profiling demonstrated a calibrated class score of 0.98 for DLGNT and copy number loss of 1p. Despite the morphologic similarities to pilocytic astrocytoma and the lack of oligodendroglial/neuronal components or leptomeningeal dissemination, the molecular profile was definitive in classifying the tumor as DLGNT. This case highlights the importance of molecular and genetic testing in the characterization of pediatric central nervous system tumors.

Phase II study of alisertib as a single agent for treating recurrent or progressive atypical teratoid/rhabdoid tumor.

BACKGROUND: Recurrent atypical teratoid/rhabdoid tumor (AT/RT) is, most often, a fatal pediatric malignancy with limited curative options. METHODS: We conducted a phase II study of Aurora kinase A inhibitor alisertib in patients aged <22 years with recurrent AT/RT. Patients received alisertib once daily (80 mg/m2 as enteric-coated tablets or 60 mg/m2 as liquid formulation) on Days 1-7 of a 21-day cycle until progressive disease (PD) occurred. Alisertib plasma concentrations were measured in cycle 1 on Days 1 (single dose) and 7 (steady state) and analyzed with noncompartmental pharmacokinetics. Trial efficacy end point was ≥10 participants with stable disease (SD) or better at 12 weeks. RESULTS: SD (n = 8) and partial response (PR) (n = 1) were observed among 30 evaluable patients. Progression-free survival (PFS) was 30.0% ± 7.9% at 6 months and 13.3% ± 5.6% at 1 year. One-year overall survival (OS) was 36.7% ± 8.4%. Two patients continued treatment for >12 months. PFS did not differ by AT/RT molecular groups. Neutropenia was the most common adverse effect (n = 23/30, 77%). The 22 patients who received liquid formulation had a higher mean maximum concentration (Cmax) of 10.1 ± 3.0 µM and faster time to Cmax (Tmax = 1.2 ± 0.7 h) than those who received tablets (Cmax = 5.7 ± 2.4 µM, Tmax = 3.4 ± 1.4 h). CONCLUSIONS: Although the study did not meet predetermined efficacy end point, single-agent alisertib was well tolerated by children with recurrent AT/RT, and SD or PR was observed in approximately a third of the patients.

Unified rhombic lip origins of group 3 and group 4 medulloblastoma.

Medulloblastoma, a malignant childhood cerebellar tumour, segregates molecularly into biologically distinct subgroups, suggesting that a personalized approach to therapy would be beneficial1. Mouse modelling and cross-species genomics have provided increasing evidence of discrete, subgroup-specific developmental origins2. However, the anatomical and cellular complexity of developing human tissues3-particularly within the rhombic lip germinal zone, which produces all glutamatergic neuronal lineages before internalization into the cerebellar nodulus-makes it difficult to validate previous inferences that were derived from studies in mice. Here we use multi-omics to resolve the origins of medulloblastoma subgroups in the developing human cerebellum. Molecular signatures encoded within a human rhombic-lip-derived lineage trajectory aligned with photoreceptor and unipolar brush cell expression profiles that are maintained in group 3 and group 4 medulloblastoma, suggesting a convergent basis. A systematic diagnostic-imaging review of a prospective institutional cohort localized the putative anatomical origins of group 3 and group 4 tumours to the nodulus. Our results connect the molecular and phenotypic features of clinically challenging medulloblastoma subgroups to their unified beginnings in the rhombic lip in the early stages of human development.

Molecular classification and outcome of children with rare CNS embryonal tumors: results from St. Jude Children's Research Hospital including the multi-center SJYC07 and SJMB03 clinical trials.

Methylation profiling has radically transformed our understanding of tumors previously called central nervous system primitive neuro-ectodermal tumors (CNS-PNET). While this marks a momentous step toward defining key differences, reclassification has thrown treatment into disarray. To shed light on response to therapy and guide clinical decision-making, we report outcomes and molecular features of children with CNS-PNETs from two multi-center risk-adapted studies (SJMB03 for patients ≥ 3 years; SJYC07 for patients < 3 years) complemented by a non-protocol institutional cohort. Seventy patients who had a histological diagnosis of CNS-PNET or CNS embryonal tumor from one of the new categories that has supplanted CNS-PNET were included. This cohort was molecularly characterized by DNA methylation profiling (n = 70), whole-exome sequencing (n = 53), RNA sequencing (n = 20), and germline sequencing (n = 28). Clinical characteristics were detailed, and treatment was divided into craniospinal irradiation (CSI)-containing (SJMB03 and SJMB03-like) and CSI-sparing therapy (SJYC07 and SJYC07-like). When the cohort was analyzed in its entirety, no differences were observed in the 5-year survival rates even when CSI-containing therapy was compared to CSI-sparing therapy. However, when analyzed by DNA methylation molecular grouping, significant survival differences were observed, and treatment particulars provided suggestions of therapeutic response. Patients with CNS neuroblastoma with FOXR2 activation (CNS-NB-FOXR2) had a 5-year event-free survival (EFS)/overall survival (OS) of 66.7% ± 19.2%/83.3% ± 15.2%, and CIC rearranged sarcoma (CNS-SARC-CIC) had a 5-year EFS/OS both of 57.1% ± 18.7% with most receiving regimens that contained radiation (focal or CSI) and multidrug chemotherapy. Patients with high-grade neuroepithelial tumor with BCOR alteration (HGNET-BCOR) had abysmal responses to upfront chemotherapy-only regimens (5-year EFS = 0%), but survival extended with salvage radiation after progression [5-year OS = 53.6% ± 20.1%]. Patients with embryonal tumor with multilayered rosettes (ETMR) or high-grade glioma/glioblastoma multiforme (HGG/GBM) did not respond favorably to any modality (5-year EFS/OS = 10.7 ± 5.8%/17.9 ± 7.2%, and 10% ± 9.0%/10% ± 9.0%, respectively). As an accompaniment, we have assembled this data onto an interactive website to allow users to probe and query the cases. By reporting on a carefully matched clinical and molecular cohort, we provide the needed insight for future clinical management.

Primary cilia control translation and the cell cycle in medulloblastoma.

The primary cilium, a signaling organelle projecting from the surface of a cell, controls cellular physiology and behavior. The presence or absence of primary cilia is a distinctive feature of a given tumor type; however, whether and how the primary cilium contributes to tumorigenesis are unknown for most tumors. Medulloblastoma (MB) is a common pediatric brain cancer comprising four groups: SHH, WNT, group 3 (G3), and group 4 (G4). From 111 cases of MB, we show that primary cilia are abundant in SHH and WNT MBs but rare in G3 and G4 MBs. Using WNT and G3 MB mouse models, we show that primary cilia promote WNT MB by facilitating translation of mRNA encoding ß-catenin, a major oncoprotein driving WNT MB, whereas cilium loss promotes G3 MB by disrupting cell cycle control and destabilizing the genome. Our findings reveal tumor type-specific ciliary functions and underlying molecular mechanisms. Moreover, we expand the function of primary cilia to translation control and reveal a molecular mechanism by which cilia regulate cell cycle progression, thereby providing new frameworks for studying cilium function in normal and pathologic conditions.

Convergent evolution and multi-wave clonal invasion in H3 K27-altered diffuse midline gliomas treated with a PDGFR inhibitor.

The majority of diffuse midline gliomas, H3 K27-altered (DMG-H3 K27-a), are infiltrating pediatric brain tumors that arise in the pons with no effective treatment. To understand how clonal evolution contributes to the tumor's invasive spread, we performed exome sequencing and SNP array profiling on 49 multi-region autopsy samples from 11 patients with pontine DMG-H3 K27-a enrolled in a phase I clinical trial of PDGFR inhibitor crenolanib. For each patient, a phylogenetic tree was constructed by testing multiple possible clonal evolution models to select the one consistent with somatic mutations and copy number variations across all tumor regions. The tree was then used to deconvolute subclonal composition and prevalence at each tumor region to study convergent evolution and invasion patterns. Somatic variants in the PI3K pathway, a late event, are enriched in our cohort, affecting 70% of patients. Convergent evolution of PI3K at distinct phylogenetic branches was detected in 40% of the patients. 24 (~ 50%) of tumor regions were occupied by subclones of mixed lineages with varying molecular ages, indicating multiple waves of invasion across the pons and extrapontine. Subclones harboring a PDGFRA amplicon, including one that amplified a PDGRFAY849C mutant allele, were detected in four patients; their presence in extrapontine tumor and normal brain samples imply their involvement in extrapontine invasion. Our study expands the current knowledge on tumor invasion patterns in DMG-H3 K27-a, which may inform the design of future clinical trials.

Comprehensive study of semi-supervised learning for DNA methylation-based supervised classification of central nervous system tumors.

BACKGROUND: Precision medicine for cancer treatment relies on an accurate pathological diagnosis. The number of known tumor classes has increased rapidly, and reliance on traditional methods of histopathologic classification alone has become unfeasible. To help reduce variability, validation costs, and standardize the histopathological diagnostic process, supervised machine learning models using DNA-methylation data have been developed for tumor classification. These methods require large labeled training data sets to obtain clinically acceptable classification accuracy. While there is abundant unlabeled epigenetic data across multiple databases, labeling pathology data for machine learning models is time-consuming and resource-intensive, especially for rare tumor types. Semi-supervised learning (SSL) approaches have been used to maximize the utility of labeled and unlabeled data for classification tasks and are effectively applied in genomics. SSL methods have not yet been explored with epigenetic data nor demonstrated beneficial to central nervous system (CNS) tumor classification. RESULTS: This paper explores the application of semi-supervised machine learning on methylation data to improve the accuracy of supervised learning models in classifying CNS tumors. We comprehensively evaluated 11 SSL methods and developed a novel combination approach that included a self-training with editing using support vector machine (SETRED-SVM) model and an L2-penalized, multinomial logistic regression model to obtain high confidence labels from a few labeled instances. Results across eight random forest and neural net models show that the pseudo-labels derived from our SSL method can significantly increase prediction accuracy for 82 CNS tumors and 9 normal controls. CONCLUSIONS: The proposed combination of semi-supervised technique and multinomial logistic regression holds the potential to leverage the abundant publicly available unlabeled methylation data effectively. Such an approach is highly beneficial in providing additional training examples, especially for scarce tumor types, to boost the prediction accuracy of supervised models.