Integrating multiomics and prior knowledge: a study of the Graphnet penalty impact.

MOTIVATION: In the field of oncology, statistical models are used for the discovery of candidate factors that influence the development of the pathology or its outcome. These statistical models can be designed in a multiblock framework to study the relationship between different multiomic data, and variable selection is often achieved by imposing constraints on the model parameters. A priori graph constraints have been used in the literature as a way to improve feature selection in the model, yielding more interpretability. However, it is still unclear how these graphs interact with the models and how they impact the feature selection. Additionally, with the availability of different graphs encoding different information, one can wonder how the choice of the graph meaningfully impacts the results obtained. RESULTS: We proposed to study the graph penalty impact on a multiblock model. Specifically, we used the SGCCA as the multiblock framework. We studied the effect of the penalty on the model using the TCGA-LGG dataset. Our findings are 3-fold. We showed that the graph penalty increases the number of selected genes from this dataset, while selecting genes already identified in other works as pertinent biomarkers in the pathology. We demonstrated that using different graphs leads to different though consistent results, but that graph density is the main factor influencing the obtained results. Finally, we showed that the graph penalty increases the performance of the survival prediction from the model-derived components and the interpretability of the results. AVAILABILITY AND IMPLEMENTATION: Source code is freely available at https://github.com/neurospin/netSGCCA.

Multimodal MRI radiomic models to predict genomic mutations in diffuse intrinsic pontine glioma with missing imaging modalities.

Purpose: Predicting H3.1, TP53, and ACVR1 mutations in DIPG could aid in the selection of therapeutic options. The contribution of clinical data and multi-modal MRI were studied for these three predictive tasks. To keep the maximum number of subjects, which is essential for a rare disease, missing data were considered. A multi-modal model was proposed, collecting all available data for each patient, without performing any imputation. Methods: A retrospective cohort of 80 patients with confirmed DIPG and at least one of the four MR modalities (T1w, T1c, T2w, and FLAIR), acquired with two different MR scanners was built. A pipeline including standardization of MR data and extraction of radiomic features within the tumor was applied. The values of radiomic features between the two MR scanners were realigned using the ComBat method. For each prediction task, the most robust features were selected based on a recursive feature elimination with cross-validation. Five different models, one based on clinical data and one per MR modality, were developed using logistic regression classifiers. The prediction of the multi-modal model was defined as the average of all possible prediction results among five for each patient. The performances of the models were compared using a leave-one-out approach. Results: The percentage of missing modalities ranged from 6 to 11% across modalities and tasks. The performance of each individual model was dependent on each specific task, with an AUC of the ROC curve ranging from 0.63 to 0.80. The multi-modal model outperformed the clinical model for each prediction tasks, thus demonstrating the added value of MRI. Furthermore, regardless of performance criteria, the multi-modal model came in the first place or second place (very close to first). In the leave-one-out approach, the prediction of H3.1 (resp. ACVR1 and TP53) mutations achieved a balanced accuracy of 87.8% (resp. 82.1 and 78.3%). Conclusion: Compared with a single modality approach, the multi-modal model combining multiple MRI modalities and clinical features was the most powerful to predict H3.1, ACVR1, and TP53 mutations and provided prediction, even in the case of missing modality. It could be proposed in the absence of a conclusive biopsy.

Object Detection Improves Tumour Segmentation in MR Images of Rare Brain Tumours.

Tumour lesion segmentation is a key step to study and characterise cancer from MR neuroradiological images. Presently, numerous deep learning segmentation architectures have been shown to perform well on the specific tumour type they are trained on (e.g., glioblastoma in brain hemispheres). However, a high performing network heavily trained on a given tumour type may perform poorly on a rare tumour type for which no labelled cases allows training or transfer learning. Yet, because some visual similarities exist nevertheless between common and rare tumours, in the lesion and around it, one may split the problem into two steps: object detection and segmentation. For each step, trained networks on common lesions could be used on rare ones following a domain adaptation scheme without extra fine-tuning. This work proposes a resilient tumour lesion delineation strategy, based on the combination of established elementary networks that achieve detection and segmentation. Our strategy allowed us to achieve robust segmentation inference on a rare tumour located in an unseen tumour context region during training. As an example of a rare tumour, Diffuse Intrinsic Pontine Glioma (DIPG), we achieve an average dice score of 0.62 without further training or network architecture adaptation.

Radiogenomics of diffuse intrinsic pontine gliomas (DIPGs): correlation of histological and biological characteristics with multimodal MRI features.

OBJECTIVES: The diffuse intrinsic pontine gliomas (DIPGs) are now defined by the type of histone H3 mutated at lysine 27. We aimed to correlate the multimodal MRI features of DIPGs, H3K27M mutant, with their histological and molecular characteristics. METHODS: Twenty-seven treatment-naïve children with histopathologically confirmed DIPG H3K27M mutant were prospectively included. MRI performed prior to biopsy included multi-b-value diffusion-weighted imaging, ASL, and dynamic susceptibility contrast (DSC) perfusion imaging. The ADC and cerebral blood flow (CBF) and blood volume (CBV) were measured at the biopsy site. We assessed quantitative histological data, including microvascular density, nuclear density, and H3K27M-positive nuclear density. Gene expression profiling was also assessed in the samples. We compared imaging and histopathological data according to histone subgroup. We correlated MRI quantitative data with histological data and gene expression. RESULTS: H3.1K27M mutated tumors showed higher ADC values (median 3151 µm2/s vs 1741 µm2/s, p = 0.003), and lower perfusion values (DSC-rCBF median 0.71 vs 1.43, p = 0.002, and DSC-rCBV median 1.00 vs 1.71, p = 0.02) than H3.3K27M ones. They had similar microvascular and nuclear density, but lower H3K27M-positive nuclear density (p = 0.007). The DSC-rCBV was positively correlated to the H3K27M-positive nuclear density (rho = 0.74, p = 0.02). ADC values were not correlated with nuclear density nor perfusion values with microvascular density. The expression of gated channel activity-related genes tended to be inversely correlated with ADC values and positively correlated with DSC perfusion. CONCLUSIONS: H3.1K27M mutated tumors have higher ADC and lower perfusion values than H3.3K27M ones, without direct correlation with microvascular or nuclear density. This may be due to tissular edema possibly related to gated channel activity-related gene expression. KEY POINTS: • H3.1K27M mutant DIPG had higher apparent diffusion coefficient (p = 0.003), lower α (p = 0.048), and lower relative cerebral blood volume (p = 0.02) than H3.3K27M mutant DIPG at their biopsy sites. • Biopsy samples obtained within the tumor's enhancing portion showed higher microvascular density (p = 0.03) than samples obtained outside the tumor's enhancing portion, but similar H3K27M-positive nuclear density (p = 0.84). • Relative cerebral blood volume measured at the biopsy site was significantly correlated with H3K27M-positive nuclear density (rho = 0.74, p = 0.02).

Clinical, molecular, and radiomic profile of gliomas with FGFR3-TACC3 fusions.

BACKGROUND: Actionable fibroblast growth factor receptor 3 (FGFR3)-transforming acidic coiled-coil protein 3 fusions (F3T3) are found in approximately 3% of gliomas, but their characteristics and prognostic significance are still poorly defined. Our goal was to characterize the clinical, radiological, and molecular profile of F3T3 positive diffuse gliomas. METHODS: We screened F3T3 fusion by real-time (RT)-PCR and FGFR3 immunohistochemistry in a large series of gliomas, characterized for main genetic alterations, histology, and clinical evolution. We performed a radiological and radiomic case control study, using an exploratory and a validation cohort. RESULTS: We screened 1162 diffuse gliomas (951 unselected cases and 211 preselected for FGFR3 protein immunopositivity), identifying 80 F3T3 positive gliomas. F3T3 was mutually exclusive with IDH mutation (P < 0.001) and EGFR amplification (P = 0.01), defining a distinct molecular cluster associated with CDK4 (P = 0.04) and MDM2 amplification (P = 0.03). F3T3 fusion was associated with longer survival for the whole series and for glioblastomas (median overall survival was 31.1 vs 19.9 mo, P = 0.02) and was an independent predictor of better outcome on multivariate analysis.F3T3 positive gliomas had specific MRI features, affecting preferentially insula and temporal lobe, and with poorly defined tumor margins. F3T3 fusion was correctly predicted by radiomics analysis on both the exploratory (area under the curve [AUC] = 0.87) and the validation MRI (AUC = 0.75) cohort. Using Cox proportional hazards models, radiomics predicted survival with a high C-index (0.75, SD 0.04), while the model combining clinical, genetic, and radiomic data showed the highest C-index (0.81, SD 0.04). CONCLUSION: F3T3 positive gliomas have distinct molecular and radiological features, and better outcome.

A genome-wide association study identifies susceptibility loci for primary central nervous system lymphoma at 6p25.3 and 3p22.1: a LOC Network study.

BACKGROUND: Primary central nervous system lymphoma (PCNSL) is a rare form of extra-nodal non-Hodgkin lymphoma. PCNSL is a distinct subtype of non-Hodgkin lymphoma, with over 95% of tumors belonging to the diffuse large B-cell lymphoma (DLBCL) group. We have conducted a genome-wide association study (GWAS) on immunocompetent patients to address the possibility that common genetic variants influence the risk of developing PCNSL. METHODS: We performed a meta-analysis of 2 new GWASs of PCNSL totaling 475 cases and 1134 controls of European ancestry. To increase genomic resolution, we imputed >10 million single nucleotide polymorphisms using the 1000 Genomes Project combined with UK10K as reference. In addition we performed a transcription factor binding disruption analysis and investigated the patterns of local chromatin by Capture Hi-C data. RESULTS: We identified independent risk loci at 3p22.1 (rs41289586, ANO10, P = 2.17 × 10-8) and 6p25.3 near EXOC2 (rs116446171, P = 1.95 x 10-13). In contrast, the lack of an association between rs41289586 and DLBCL suggests distinct germline predisposition to PCNSL and DLBCL. We found looping chromatin interactions between noncoding regions at 6p25.3 (rs11646171) with the IRF4 promoter and at 8q24.21 (rs13254990) with the MYC promoter, both genes with strong relevance to B-cell tumorigenesis. CONCLUSION: To our knowledge this is the first study providing insight into the genetic predisposition to PCNSL. Our findings represent an important step in defining the contribution of common genetic variation to the risk of developing PCNSL.

Transcriptomic and epigenetic profiling of 'diffuse midline gliomas, H3 K27M-mutant' discriminate two subgroups based on the type of histone H3 mutated and not supratentorial or infratentorial location.

Diffuse midline glioma (DMG), H3 K27M-mutant, is a new entity in the updated WHO classification grouping together diffuse intrinsic pontine gliomas and infiltrating glial neoplasms of the midline harboring the same canonical mutation at the Lysine 27 of the histones H3 tail.Two hundred and fifteen patients younger than 18 years old with centrally-reviewed pediatric high-grade gliomas (pHGG) were included in this study. Comprehensive transcriptomic (n = 140) and methylation (n = 80) profiling was performed depending on the material available, in order to assess the biological uniqueness of this new entity compared to other midline and hemispheric pHGG.Tumor classification based on gene expression (GE) data highlighted the similarity of K27M DMG independently of their location along the midline. T-distributed Stochastic Neighbor Embedding (tSNE) analysis of methylation profiling confirms the discrimination of DMG from other well defined supratentorial tumor subgroups. Patients with diffuse intrinsic pontine gliomas (DIPG) and thalamic DMG exhibited a similarly poor prognosis (11.1 and 10.8 months median overall survival, respectively). Interestingly, H3.1-K27M and H3.3-K27M primary tumor samples could be distinguished based both on their GE and DNA methylation profiles, suggesting that they might arise from a different precursor or from a different epigenetic reorganization.These differences in DNA methylation profiles were conserved in glioma stem-like cell culture models of DIPG which mimicked their corresponding primary tumor. ChIP-seq profiling of H3K27me3 in these models indicate that H3.3-K27M mutated DIPG stem cells exhibit higher levels of H3K27 trimethylation which are correlated with fewer genes expressed by RNAseq. When considering the global distribution of the H3K27me3 mark, we observed that intergenic regions were more trimethylated in the H3.3-K27M mutated cells compared to the H3.1-K27M mutated ones.H3 K27M-mutant DMG represent a homogenous group of neoplasms compared to other pediatric gliomas that could be further separated based on the type of histone H3 variant mutated and their respective epigenetic landscapes. As these characteristics drive different phenotypes, these findings may have important implication for the design of future trials in these specific types of neoplasms.

Computation of reliable textural indices from multimodal brain MRI: suggestions based on a study of patients with diffuse intrinsic pontine glioma.

Few methodological studies regarding widely used textural indices robustness in MRI have been reported. In this context, this study aims to propose some rules to compute reliable textural indices from multimodal 3D brain MRI. Diagnosis and post-biopsy MR scans including T1, post-contrast T1, T2 and FLAIR images from thirty children with diffuse intrinsic pontine glioma (DIPG) were considered. The hybrid white stripe method was adapted to standardize MR intensities. Sixty textural indices were then computed for each modality in different regions of interest (ROI), including tumor and white matter (WM). Three types of intensity binning were compared [Formula: see text]: constant bin width and relative bounds; [Formula: see text] constant number of bins and relative bounds; [Formula: see text] constant number of bins and absolute bounds. The impact of the volume of the region was also tested within the WM. First, the mean Hellinger distance between patient-based intensity distributions decreased by a factor greater than 10 in WM and greater than 2.5 in gray matter after standardization. Regarding the binning strategy, the ranking of patients was highly correlated for 188/240 features when comparing [Formula: see text] with [Formula: see text], but for only 20 when comparing [Formula: see text] with [Formula: see text], and nine when comparing [Formula: see text] with [Formula: see text]. Furthermore, when using [Formula: see text] or [Formula: see text] texture indices reflected tumor heterogeneity as assessed visually by experts. Last, 41 features presented statistically significant differences between contralateral WM regions when ROI size slightly varies across patients, and none when using ROI of the same size. For regions with similar size, 224 features were significantly different between WM and tumor. Valuable information from texture indices can be biased by methodological choices. Recommendations are to standardize intensities in MR brain volumes, to use intensity binning with constant bin width, and to define regions with the same volumes to get reliable textural indices.

A Postreconstruction Harmonization Method for Multicenter Radiomic Studies in PET.

Several reports have shown that radiomic features are affected by acquisition and reconstruction parameters, thus hampering multicenter studies. We propose a method that, by removing the center effect while preserving patient-specific effects, standardizes features measured from PET images obtained using different imaging protocols. Methods: Pretreatment 18F-FDG PET images of patients with breast cancer were included. In one nuclear medicine department (department A), 63 patients were scanned on a time-of-flight PET/CT scanner, and 16 lesions were triple-negative (TN). In another nuclear medicine department (department B), 74 patients underwent PET/CT on a different brand of scanner and a different reconstruction protocol, and 15 lesions were TN. The images from department A were smoothed using a gaussian filter to mimic data from a third department (department A-S). The primary lesion was segmented to obtain a lesion volume of interest (VOI), and a spheric VOI was set in healthy liver tissue. Three SUVs and 6 textural features were computed in all VOIs. A harmonization method initially described for genomic data was used to estimate the department effect based on the observed feature values. Feature distributions in each department were compared before and after harmonization. Results: In healthy liver tissue, the distributions significantly differed for 4 of 9 features between departments A and B and for 6 of 9 between departments A and A-S (P < 0.05, Wilcoxon test). After harmonization, none of the 9 feature distributions significantly differed between 2 departments (P > 0.1). The same trend was observed in lesions, with a realignment of feature distributions between the departments after harmonization. Identification of TN lesions was largely enhanced after harmonization when the cutoffs were determined on data from one department and applied to data from the other department. Conclusion: The proposed harmonization method is efficient at removing the multicenter effect for textural features and SUVs. The method is easy to use, retains biologic variations not related to a center effect, and does not require any feature recalculation. Such harmonization allows for multicenter studies and for external validation of radiomic models or cutoffs and should facilitate the use of radiomic models in clinical practice.

Histone H3F3A and HIST1H3B K27M mutations define two subgroups of diffuse intrinsic pontine gliomas with different prognosis and phenotypes.

Diffuse intrinsic pontine glioma (DIPG) is the most severe paediatric solid tumour, with no significant therapeutic progress made in the past 50 years. Recent studies suggest that diffuse midline glioma, H3-K27M mutant, may comprise more than one biological entity. The aim of the study was to determine the clinical and biological variables that most impact their prognosis. Ninety-one patients with classically defined DIPG underwent a systematic stereotactic biopsy and were included in this observational retrospective study. Histone H3 genes mutations were assessed by immunochemistry and direct sequencing, whilst global gene expression profiling and chromosomal imbalances were determined by microarrays. A full description of the MRI findings at diagnosis and at relapse was integrated with the molecular profiling data and clinical outcome. All DIPG but one were found to harbour either a somatic H3-K27M mutation and/or loss of H3K27 trimethylation. We also discovered a novel K27M mutation in HIST2H3C, and a lysine-to-isoleucine substitution (K27I) in H3F3A, also creating a loss of trimethylation. Patients with tumours harbouring a K27M mutation in H3.3 (H3F3A) did not respond clinically to radiotherapy as well, relapsed significantly earlier and exhibited more metastatic recurrences than those in H3.1 (HIST1H3B/C). H3.3-K27M-mutated DIPG have a proneural/oligodendroglial phenotype and a pro-metastatic gene expression signature with PDGFRA activation, while H3.1-K27M-mutated tumours exhibit a mesenchymal/astrocytic phenotype and a pro-angiogenic/hypoxic signature supported by expression profiling and radiological findings. H3K27 alterations appear as the founding event in DIPG and the mutations in the two main histone H3 variants drive two distinct oncogenic programmes with potential specific therapeutic targets.

Preclinical evaluation of dasatinib alone and in combination with cabozantinib for the treatment of diffuse intrinsic pontine glioma.

BACKGROUND: Platelet-derived growth factor receptor A is altered by amplification and/or mutation in diffuse intrinsic pontine glioma (DIPG). We explored in vitro on new DIPG models the efficacy of dasatinib, a multi-tyrosine kinase inhibitor targeting this receptor. METHODS: Gene expression profiles were generated from 41 DIPGs biopsied at diagnosis and compared with the signature associated with sensitivity/resistance to dasatinib. A panel of 12 new DIPG cell lines were established from biopsy at diagnosis, serially passaged, and characterized by gene expression analyses. Effects of dasatinib (1-10 µM) on proliferation, invasion, and cytotoxicity were determined on 4 of these cell lines using live-cell imaging and flow cytometry assays. Downstream signaling and receptor tyrosine kinases (RTKs) were assessed by western blot and phospho-RTK array. The effect of the combination with the c-Met inhibitor cabozantinib was studied on cellular growth and invasion analyzed by the Chou-Talaly method. RESULTS: DIPG primary tumors and cell lines exhibited the gene expression signature of sensitivity to dasatinib. Dasatinib reduced proliferation (half-maximal inhibitory concentration = 10-100 nM) and invasion (30%-60% reduction) at 100 nM in 4/4 cultures and induced apoptosis in 1 of 4 DIPG cell lines. Activity of downstream effectors of dasatinib targets including activin receptor 1 was strongly reduced. Since multiple RTKs were activated simultaneously in DIPG cell lines, including c-Met, which can be also amplified in DIPG, the benefit of the combination of dasatinib with cabozantinib was explored for its synergistic effects on proliferation and migration/invasion in these cell lines. CONCLUSION: Dasatinib exhibits antitumor effects in vitro that could be increased by the combination with another RTK inhibitor targeting c-Met.

Recurrent activating ACVR1 mutations in diffuse intrinsic pontine glioma.

Diffuse intrinsic pontine gliomas (DIPGs) are highly infiltrative malignant glial neoplasms of the ventral pons that, due to their location within the brain, are unsuitable for surgical resection and consequently have a universally dismal clinical outcome. The median survival time is 9-12 months, with neither chemotherapeutic nor targeted agents showing substantial survival benefit in clinical trials in children with these tumors. We report the identification of recurrent activating mutations in the ACVR1 gene, which encodes a type I activin receptor serine/threonine kinase, in 21% of DIPG samples. Strikingly, these somatic mutations (encoding p.Arg206His, p.Arg258Gly, p.Gly328Glu, p.Gly328Val, p.Gly328Trp and p.Gly356Asp substitutions) have not been reported previously in cancer but are identical to mutations found in the germ line of individuals with the congenital childhood developmental disorder fibrodysplasia ossificans progressiva (FOP) and have been shown to constitutively activate the BMP-TGF-ß signaling pathway. These mutations represent new targets for therapeutic intervention in this otherwise incurable disease.

Variable selection for generalized canonical correlation analysis.

Regularized generalized canonical correlation analysis (RGCCA) is a generalization of regularized canonical correlation analysis to 3 or more sets of variables. RGCCA is a component-based approach which aims to study the relationships between several sets of variables. The quality and interpretability of the RGCCA components are likely to be affected by the usefulness and relevance of the variables in each block. Therefore, it is an important issue to identify within each block which subsets of significant variables are active in the relationships between blocks. In this paper, RGCCA is extended to address the issue of variable selection. Specifically, sparse generalized canonical correlation analysis (SGCCA) is proposed to combine RGCCA with an [Formula: see text]-penalty in a unified framework. Within this framework, blocks are not necessarily fully connected, which makes SGCCA a flexible method for analyzing a wide variety of practical problems. Finally, the versatility and usefulness of SGCCA are illustrated on a simulated dataset and on a 3-block dataset which combine gene expression, comparative genomic hybridization, and a qualitative phenotype measured on a set of 53 children with glioma. SGCCA is available on CRAN as part of the RGCCA package.

Mesenchymal transition and PDGFRA amplification/mutation are key distinct oncogenic events in pediatric diffuse intrinsic pontine gliomas.

Diffuse intrinsic pontine glioma (DIPG) is one of the most frequent malignant pediatric brain tumor and its prognosis is universaly fatal. No significant improvement has been made in last thirty years over the standard treatment with radiotherapy. To address the paucity of understanding of DIPGs, we have carried out integrated molecular profiling of a large series of samples obtained with stereotactic biopsy at diagnosis. While chromosomal imbalances did not distinguish DIPG and supratentorial tumors on CGHarrays, gene expression profiling revealed clear differences between them, with brainstem gliomas resembling midline/thalamic tumours, indicating a closely-related origin. Two distinct subgroups of DIPG were identified. The first subgroup displayed mesenchymal and pro-angiogenic characteristics, with stem cell markers enrichment consistent with the possibility to grow tumor stem cells from these biopsies. The other subgroup displayed oligodendroglial features, and appeared largely driven by PDGFRA, in particular through amplification and/or novel missense mutations in the extracellular domain. Patients in this later group had a significantly worse outcome with an hazard ratio for early deaths, ie before 10 months, 8 fold greater that the ones in the other subgroup (p = 0.041, Cox regression model). The worse outcome of patients with the oligodendroglial type of tumors was confirmed on a series of 55 paraffin-embedded biopsy samples at diagnosis (median OS of 7.73 versus 12.37 months, p = 0.045, log-rank test). Two distinct transcriptional subclasses of DIPG with specific genomic alterations can be defined at diagnosis by oligodendroglial differentiation or mesenchymal transition, respectively. Classifying these tumors by signal transduction pathway activation and by mutation in pathway member genes may be particularily valuable for the development of targeted therapies.

Critical oncogenic mutations in newly diagnosed pediatric diffuse intrinsic pontine glioma.

Diffuse intrinsic pontine gliomas (DIPG) can not be cured with current treatment modalities. Targeted therapy in this disease would benefit from advanced technologies detecting relevant drugable mutations. Twenty patients with classic newly diagnosed DIPG underwent stereotactic biopsies and were analyzed for the presence of 983 different mutations in 115 oncogenes and tumor-suppressor genes using OncoMap, a mass spectrometric method of allele detection. Our results identified oncogenic mutations in TP53 (40%), PI3KCA (15%), and ATM/MPL (5%) while none were identified in a large number of other genes commonly mutated in malignant gliomas. The identification of oncogenic mutations in the PI3K pathway offers the potential of a therapeutic target at initial diagnosis in this devastating disease.

Neuronal differentiation distinguishes supratentorial and infratentorial childhood ependymomas.

Ependymomas are glial neoplasms occurring in any location throughout the central nervous system and supposedly are derived from radial glia cells. Recent data suggest that these tumors may have different biological and clinical behaviors according to their location. Pediatric supratentorial and infratentorial ependymoma (SE and IE) were compared with respect to clinical and radiological parameters and immunohistochemistry (IHC). Neuronal markers were specifically assessed by IHC and quantitative PCR (qPCR). No single morphological or radiological characteristic was associated with location or any neuronal marker. However, there was a significant overexpression of neuronal markers in SE compared with IE: neurofilament light polypeptide 70 (NEFL)-positive tumor cells were found in 23 of 34 SE and in only 4 of 32 IE (P < .001). Among SE, 10 of 34 exhibited high expression of NEFL, defined as more than 5% positive cells. qPCR confirmed the upregulation of neuronal markers (NEFL, LHX2, FOXG1, TLX1, and NPTXR) in SE compared with IE. In addition, strong NEFL expression in SE was correlated with better progression-free survival (P = .007). Our results support the distinction of SE and IE. SEs are characterized by neuronal differentiation, which seems to be associated with better prognosis.