t-SNE Analysis Research Articles

This paper examines the link between CNS tumor biology and heterogeneity and the use of genome-wide DNA methylation profiling as a clinical diagnostic platform. CNS tumors are the most common solid tumors in children, and their prognosis remains poor. This study retrospectively analyzed pediatric patients with CNS embryonal tumors in Hong Kong between 1999 and 2017, using data from the territory-wide registry and available formalin-fixed paraffin-embedded tumor tissue. After processing archival tumor tissue via DNA extraction, quantification, and methylation profiling, the data were analyzed by using the web-based DKFZ classifier (Molecular Neuropathology (MNP) 2.0 v11b4) and t-SNE analysis. Methylation profiles were deemed informative in 85 samples. Epigenetic data allowed molecular subgrouping and confirmed diagnosis in 65 samples, verified histologic diagnosis in 8, and suggested an alternative diagnosis in 12. This study demonstrates the potential of DNA methylation profiling in characterizing pediatric CNS embryonal tumors in a large cohort from Hong Kong, which should enable regional and international collaboration in future pediatric neuro-oncology research.

BackgroundExtracellular vesicles (EVs) originating from the central nervous system (CNS) can enter the blood stream and carry molecules characteristic of disease states. Therefore, circulating CNS-derived EVs have the potential to serve as liquid-biopsy markers for early diagnosis and follow-up of neurodegenerative diseases and brain tumors. Monitoring and profiling of CNS-derived EVs using multiparametric analysis would be a major advance for biomarker as well as basic research. Here, we explored the performance of a multiplex bead-based flow-cytometry assay (EV Neuro) for semi-quantitative detection of CNS-derived EVs in body fluids.MethodsEVs were separated from culture of glioblastoma cell lines (LN18, LN229, NCH82) and primary human astrocytes and measured at different input amounts in the MACSPlex EV Kit Neuro, human. In addition, EVs were separated from blood samples of small cohorts of glioblastoma (GB), multiple sclerosis (MS) and Alzheimer’s disease patients as well as healthy controls (HC) and subjected to the EV Neuro assay. To determine statistically significant differences between relative marker signal intensities, an unpaired samples t-test or Wilcoxon rank sum test were computed. Data were subjected to tSNE, heatmap clustering, and correlation analysis to further explore the relationships between disease state and EV Neuro data.ResultsGlioblastoma cell lines and primary human astrocytes showed distinct EV profiles. Signal intensities were increasing with higher EV input. Data normalization improved identification of markers that deviate from a common profile. Overall, patient blood-derived EV marker profiles were constant, but individual EV populations were significantly increased in disease compared to healthy controls, e.g. CD36+EVs in glioblastoma and GALC+EVs in multiple sclerosis. tSNE and heatmap clustering analysis separated GB patients from HC, but not MS patients from HC. Correlation analysis revealed a potential association of CD107a+EVs with neurofilament levels in blood of MS patients and HC.ConclusionsThe semi-quantitative EV Neuro assay demonstrated its utility for EV profiling in complex samples. However, reliable statistical results in biomarker studies require large sample cohorts and high effect sizes. Nonetheless, this exploratory trial confirmed the feasibility of discovering EV-associated biomarkers and monitoring circulating EV profiles in CNS diseases using the EV Neuro assay.-KwzeJiLkD23APPgby_EgVVideo

Visual hallucinations in Parkinson's disease can be viewed from a systems-level perspective, whereby dysfunctional communication between brain networks responsible for perception predisposes a person to hallucinate. To this end, abnormal functional interactions between higher-order and primary sensory networks have been implicated in the pathophysiology of visual hallucinations in Parkinson's disease, however the precise signatures remain to be determined. Dimensionality reduction techniques offer a novel means for simplifying the interpretation of multidimensional brain imaging data, identifying hierarchical patterns in the data that are driven by both within- and between-functional network changes. Here, we applied two complementary non-linear dimensionality reduction techniques-diffusion-map embedding and t-distributed stochastic neighbour embedding (t-SNE)-to resting state functional MRI data, in order to characterize the altered functional hierarchy associated with susceptibility to visual hallucinations. Our study involved 77 people with Parkinson's disease (31 with hallucinations; 46 without hallucinations) and 19 age-matched healthy control subjects. In patients with visual hallucinations, we found compression of the unimodal-heteromodal gradient consistent with increased functional integration between sensory and higher order networks. This was mirrored in a traditional functional connectivity analysis, which showed increased connectivity between the visual and default mode networks in the hallucinating group. Together, these results suggest a route by which higher-order regions may have excessive influence over earlier sensory processes, as proposed by theoretical models of hallucinations across disorders. By contrast, the t-SNE analysis identified distinct alterations in prefrontal regions, suggesting an additional layer of complexity in the functional brain network abnormalities implicated in hallucinations, which was not apparent in traditional functional connectivity analyses. Together, the results confirm abnormal brain organization associated with the hallucinating phenotype in Parkinson's disease and highlight the utility of applying convergent dimensionality reduction techniques to investigate complex clinical symptoms. In addition, the patterns we describe in Parkinson's disease converge with those seen in other conditions, suggesting that reduced hierarchical differentiation across sensory-perceptual systems may be a common transdiagnostic vulnerability in neuropsychiatric disorders with perceptual disturbances.

Abstract Chronic alcohol consumption is associated with significant mortality and morbidity in people affected by Mycobacterium tuberculosis (Mtb) infection. Our Lab previously reported that alcohol diet increased mortality in young mice infected with Mtb in an Interferon-α (IFN-α) dependent manner. We also found CD11b+Ly6g+ cells are major source for IFN-α. In the current study, using single cell RNA sequencing, we further characterized IFN-α producing CD11b+Ly6g+ cells in the lungs of Mtb infected alcohol-diet fed mice. Our analysis revealed ‘6’ distinct clusters in CD11b+Ly6g+ cell population based on gene signatures. We identified cluster ‘6’, through gene-set scoring and expression of interferon related genes such as IRF1, IRF5 to be responsible for type I interferon production. Pseudotime analysis along with KEGG enrichment, revealed transcriptome of cluster ‘6’ to be in a separate trajectory compared to the core of the cluster, suggesting a differentiated phenotype. Similarly, we identified CD69 and CD74 as a surface marker for this unique cell population which was subsequently verified with t-sne analysis using flow cytometry. We are currently characterizing the cellular mechanisms regulating the expansion of these cells in Mtb infected alcohol-diet fed mice to be able to target them for effective early Interferon-α neutralization.

To evaluate the benefits of SARS-CoV-2 vaccination in cancer patients it is relevant to understand the adaptive immune response elicited after vaccination. Patients affected by hematologic malignancies are frequently immune-compromised and show a decreased seroconversion rate compared to other cancer patients or controls. Therefore, vaccine-induced cellular immune responses in these patients might have an important protective role and need a detailed evaluation. Certain T cell subtypes (CD4, CD8, Tfh, γδT), including cell functionality as indicated by cytokine secretion (IFN, TNF) and expression of activation markers (CD69, CD154) were assessed via multi-parameter flow cytometry in hematologic malignancy patients (N=12) and healthy controls (N=12) after a second SARS-CoV-2 vaccine dose. The PBMC of post-vaccination samples were stimulated with a spike-peptide pool (S-Peptides) of SARS-CoV-2, with CD3/CD28, with a pool of peptides from the cytomegalovirus, Epstein-Barr virus and influenza A virus (CEF-Peptides) or left unstimulated. Furthermore, the concentration of spike-specific antibodies has been analyzed in patients. Our results indicate that hematologic malignancy patients developed a robust cellular immune response to SARS-CoV-2 vaccination comparable to that of healthy controls, and for certain T cell subtypes even higher. The most reactive T cells to SARS-CoV-2 spike peptides belonged to the CD4 and Tfh cell compartment, being median (IQR), 3.39 (1.41-5.92) and 2.12 (0.55-4.14) as a percentage of IFN- and TNF-producing Tfh cells in patients. In this regard, the immunomodulatory treatment of patients before the vaccination period seems important as it was strongly associated with a higher percentage of activated CD4 and Tfh cells. SARS-CoV-2- and CEF-specific T cell responses significantly correlated with each other. Compared to lymphoma patients, myeloma patients had an increased percentage of SARS-CoV-2-specific Tfh cells. T-SNE analysis revealed higher frequencies of γδT cells in patients compared to controls, especially in myeloma patients. In general, after vaccination, SARS-CoV-2-specific T cells were also detectable in patients without seroconversion. Hematologic malignancy patients are capable of developing a SARS-CoV-2-specific CD4 and Tfh cellular immune response after vaccination, and certain immunomodulatory therapies in the period before vaccination might increase the antigen-specific immune response. A proper response to recall antigens (e.g., CEF-Peptides) reflects immune cellular functionality and might be predictive for generating a newly induced antigen-specific immune response as is expected after SARS-CoV-2 vaccination.

Correlated and individual feature learning with contrast-enhanced MR for malignancy characterization of hepatocellular carcinoma

Gastrointestinal (GI) cancer accounts for one in four cancer cases and one in three cancer-related deaths globally. A deeper understanding of cancer development mechanisms can be applied to cancer medicine. Comprehensive sequencing applications have revealed the genomic landscapes of the common types of human cancer, and proteomics technology has identified protein targets and signalling pathways related to cancer growth and progression. This study aimed to explore the functional proteomic profiles of four major types of GI tract cancer based on The Cancer Proteome Atlas (TCPA). We provided an overview of functional proteomic heterogeneity by performing several approaches, including principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), t-stochastic neighbour embedding (t-SNE) analysis, and hierarchical clustering analysis in oesophageal carcinoma (ESCA), stomach adenocarcinoma (STAD), colon adenocarcinoma (COAD), and rectum adenocarcinoma (READ) tumours, to gain a system-wide understanding of the four types of GI cancer. The feature selection approach, mutual information feature selection (MIFS) method, was conducted to screen candidate protein signature subsets to better distinguish different cancer types. The potential clinical implications of candidate proteins in terms of tumour progression and prognosis were also evaluated based on TCPA and The Cancer Genome Atlas (TCGA) databases. The results suggested that functional proteomic profiling can identify different patterns among the four types of GI cancers and provide candidate proteins for clinical diagnosis and prognosis evaluation. We also highlighted the application of feature selection approaches in high-dimensional biological data analysis. Overall, this study could improve the understanding of the complexity of cancer phenotypes and genotypes and thus be applied to cancer medicine.

Plexiform neurofibromas can transform into atypical neurofibromas (ANF) and then further progress to aggressive malignant peripheral nerve sheath tumors (MPNST). ANF have been described to harbor distinct histological features and frequent loss of CDKN2A/B. However, histological evaluation may be rater-dependent, and detailed knowledge about the molecular mechanisms of malignant transformation is scarce. In general, malignant transformation can be accompanied by significant epigenetic changes, and global DNA methylation profiling is able to differentiate relevant tumor subgroups. Therefore, epigenetic profiling might provide a valuable tool to distinguish and characterize ANF with differing extent of histopathological atypia from neurofibromas and MPNST. We investigated 40 tumors histologically diagnosed as ANF and compared their global methylation profile to other peripheral nerve sheath tumors. Unsupervised class discovery and t-SNE analysis indicated that 36/40 ANF cluster with benign peripheral nerve sheath tumors with clear separation from MPNST. 21 ANF formed a molecularly distinct cluster in proximity to schwannomas. Tumors in this cluster had a frequent heterozygous or homozygous loss of CDKN2A/B and significantly more lymphocyte infiltration than MPNST, schwannomas, and NF. Few ANF clustered closely with neurofibromas, schwannomas, or MPNST, raising the question, whether diagnosis based on histological features alone might pose a risk to both over- and underestimate the aggressiveness of these lesions. Our data suggest that ANF with varying histological morphology show distinct epigenetic similarities and cluster in proximity to benign peripheral nerve sheath tumor entities. Future investigations should pay special respect to correlating this methylation pattern to clinical outcomes.

Artificial intelligence (AI) technologies are used in many dimensions of our lives, including education. Motivated by the increasing use of AI technologies and the current state of the art, this study examines research on AI from the perspective of online distance education. Following a systematic review protocol and using data mining and analytics approaches, the study examines a total of 276 publications. Accordingly, time trend analysis increases steadily with a peak in recent years, and China, India, and the United States are the leading countries in research on AI in online learning and distance education. Computer science and engineering are the research areas that make the most of the contribution, followed by social sciences. t-SNE analysis reveals three dominant clusters showing thematic tendencies, which are as follows: (1) how AI technologies are used in online teaching and learning processes, (2) how algorithms are used for the recognition, identification, and prediction of students’ behaviors, and (3) adaptive and personalized learning empowered through artificial intelligence technologies. Additionally, the text mining and social network analysis identified three broad research themes, which are (1) educational data mining, learning analytics, and artificial intelligence for adaptive and personalized learning; (2) algorithmic online educational spaces, ethics, and human agency; and (3) online learning through detection, identification, recognition, and prediction.

BackgroundGliomas with FGFR3::TACC3 fusion mainly occur in adults, display pathological features of glioblastomas (GB) and are usually classified as glioblastoma, IDH-wildtype. However, cases demonstrating pathological features of low-grade glioma (LGG) lead to difficulties in classification and clinical management. We report a series of 8 GB and 14 LGG with FGFR3:TACC3 fusion in order to better characterize them.MethodsCentralized pathological examination, search for TERT promoter mutation and DNA-methylation profiling were performed in all cases. Search for prognostic factors was done by the Kaplan–Meir method.ResultsTERT promoter mutation was recorded in all GB and 6/14 LGG. Among the 7 cases with a methylation score > 0.9 in the classifier (v12.5), 2 were classified as glioblastoma, 4 as ganglioglioma (GG) and 1 as dysembryoplastic neuroepithelial tumor (DNET). t-SNE analysis showed that the 22 cases clustered into three groups: one included 12 cases close to glioblastoma, IDH-wildtype methylation class (MC), 5 cases each clustered with GG or DNET MC but none with PLNTY MC. Unsupervised clustering analysis revealed four groups, two of them being clearly distinct: 5 cases shared age (< 40), pathological features of LGG, lack of TERT promoter mutation, FGFR3(Exon 17)::TACC3(Exon 10) fusion type and LGG MC. In contrast, 4 cases shared age (> 40), pathological features of glioblastoma, and were TERT-mutated. Relevant factors associated with a better prognosis were age < 40 and lack of TERT promoter mutation.ConclusionAmong gliomas with FGFR3::TACC3 fusion, age, TERT promoter mutation, pathological features, DNA-methylation profiling and fusion subtype are of interest to determine patients’ risk.

Predictive biomarkers for immune checkpoint inhibitors (ICIs), e.g., programmed death ligand-1 (PD-L1) tumor proportional score (TPS), remain limited in clinical applications. Predictive biomarkers that require invasive tumor biopsy procedures are practically challenging especially when longitudinal follow-up is required. Clinical utility of tissue-based PD-L1 TPS also becomes diluted when ICI is combined with chemotherapies. Peripheral single T-cell dynamic polyfunctionality profiling offers the opportunity to reveal rare T-cell subpopulations that are polyfunctional and responsible for the underlying ICI treatment molecular response that bulk biological assays cannot achieve. Here, we evaluated a novel live single-cell functional liquid biopsy cytokine profiling platform, IsoLight, as a potential predictive biomarker to track ICI treatment response and clinical outcomes in non-small cell lung cancer (NSCLC). Peripheral blood mononuclear cell samples of 10 healthy donors and 10 NSCLC patients undergoing ICI-based therapies were collected longitudinally pre-/post-ICI treatment after ≥2 cycles under institutional review board (IRB)-approved protocols. Cancer blood samples were collected from unresectable advanced stage (III-IV) NSCLC patients. Clinical course and treatment response and survival outcomes were extracted from electronic health records, with treatment response assessed by treating oncologists based on RECIST. CD4+ and CD8+ T-cells were enriched magnetically and analyzed on the IsoLight platform. Single T-cells were captured in microchambers on IsoCode chips for proteomic immune cytokines profiling. Functional polyfunctionality data from 55,775 single cells were analyzed with IsoSpeak software, 2D- and 3D-t-distributed stochastic neighbor embedding (t-SNE) analysis, kappa coefficient, and Kaplan-Meier survival plots. P values ≤0.05 is considered statistically significant. Pre-treatment baseline polyfunctionality profiles could not differentiate NSCLC patients from healthy subjects, and could not differentiate ICI responders from non-responders. We found a statistically significant difference between responders and non-responders in CD8+ T-cells' changes in overall polyfunctionality (ΔPolyFx) (P=0.01) and polyfunctional strength index (ΔPSI) (P=0.006) in our dynamic pre-/post-treatment single cell measurements, both performing better than PD-L1 TPS alone (P=0.08). In the 3D-t-SNE analysis, subpopulations of post-treatment CD8+ T-cells in ICI responders displayed distinct immune cytokine profiles from those in pre-treatment cells. CD8+ T-cells ΔPolyFx and ΔPSI scores performed better than PD-L1 TPS in ICI response correlation. Moreover, combined PD-L1 strong TPS and ΔPSI >15 scores strongly correlated with early ICI response with a robust kappa coefficient of 1.0 (P=0.003), which was previously statistically established to indicate a perfect agreement between the prediction and actual response status. Interestingly, high CD4+ T-cells ΔPSI >5 was found to correlate with a strong trend of improved progression-free survival (3.9-fold) (10.8 vs. 2.8 months; P=0.07) and overall survival (3-fold) (34.5 vs. 11.5 months; P=0.09) in ICI-treated patients. Our study nominates single peripheral T-cell polyfunctionality dynamics analysis to be a promising liquid biopsy platform to determine potential ICI predictive biomarker in NSCLC. It warrants further studies in larger prospective cohorts to validate the clinical utilities and to further optimize cancer immunotherapy.

Ovarian cancer is a lethal female reproductive system malignancy. However, the physiological roles of ferroptosis in ovarian cancer remains unclear. In this study, biological information databases were screened to characterize and examine the differentially expressed ferroptosis-related genes between ovarian cancer and normal ovarian tissue, and to further investigate a novel risk signature for predicting the prognosis of ovarian cancer. Molecular and clinical data were retrieved from The Cancer Genome Atlas (TCGA) database. Based on these data, we identified differentially expressed ferroptosis-related genes, and construct a multigene risk signature by least absolute shrinkage and celection operator (LASSO) Cox regression to predict the prognosis of ovarian cancer. Univariate and multivariate Cox regression analysis were used to verify the prognostic value of the signature. We constructed a risk signature for ovarian cancer based on differentially expressed ferroptosis-related genes between normal ovarian samples and ovarian cancer samples. Referring to median risk score, patients were divided into high-risk group and low-risk group. We performed Cox regression analysis, principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE) analysis, Kaplan-Meier Survival analysis and receiver operating characteristic (ROC) curve to verify the accuracy of the predicted value of the risk signature. The overall survival rates in low-risk group was significantly higher than that in high-risk group. In addition, the area under the curve (AUC) of the ROC curve reached 0.684 at 1 year, 0.682 at 2 years and 0.661 at 3 years. Functional analysis indicated differentially expressed ferroptosis-related genes were enriched in immune-related cells. The ferroptosis-related genes signature could predict the prognosis of ovarian cancer. These genes might be potential therapeutic targets.

BackgroundAccurate classification of sites of interest on prostate-specific membrane antigen (PSMA) positron emission tomography (PET) images is an important diagnostic requirement for the differentiation of prostate cancer (PCa) from foci of physiologic uptake. We developed a deep learning and radiomics framework to perform lesion-level and patient-level classification on PSMA PET images of patients with PCa.MethodsThis was an IRB-approved, HIPAA-compliant, retrospective study. Lesions on [18F]DCFPyL PET/CT scans were assigned to PSMA reporting and data system (PSMA-RADS) categories and randomly partitioned into training, validation, and test sets. The framework extracted image features, radiomic features, and tissue type information from a cropped PET image slice containing a lesion and performed PSMA-RADS and PCa classification. Performance was evaluated by assessing the area under the receiver operating characteristic curve (AUROC). A t-distributed stochastic neighbor embedding (t-SNE) analysis was performed. Confidence and probability scores were measured. Statistical significance was determined using a two-tailed t test.ResultsPSMA PET scans from 267 men with PCa had 3794 lesions assigned to PSMA-RADS categories. The framework yielded AUROC values of 0.87 and 0.90 for lesion-level and patient-level PSMA-RADS classification, respectively, on the test set. The framework yielded AUROC values of 0.92 and 0.85 for lesion-level and patient-level PCa classification, respectively, on the test set. A t-SNE analysis revealed learned relationships between the PSMA-RADS categories and disease findings. Mean confidence scores reflected the expected accuracy and were significantly higher for correct predictions than for incorrect predictions (P < 0.05). Measured probability scores reflected the likelihood of PCa consistent with the PSMA-RADS framework.ConclusionThe framework provided lesion-level and patient-level PSMA-RADS and PCa classification on PSMA PET images. The framework was interpretable and provided confidence and probability scores that may assist physicians in making more informed clinical decisions.

To develop a deep-learning system for the automatic identification of triple-negative breast cancer (TNBC) solely from ultrasound images. A total of 145 patients and 831 images were retrospectively enrolled at Peking Union College Hospital from April 2018 to March 2019. Ultrasound images and clinical information were collected accordingly. Molecular subtypes were determined from immunohistochemical (IHC) results. A CNN with VGG-based architecture was then used to predict TNBC. The model’s performance was evaluated using randomized k-fold stratified cross-validation. A t-SNE analysis and saliency maps were used for model visualization. TNBC was identified in 16 of 145 (11.03%) patients. One hundred fifteen (80%) patients, 15 (10%) patients, and 15 (10%) patients formed the train, validation, and test set respectively. The deep learning system exhibits good efficacy, with an AUC of 0.86 (95% CI: 0.64, 0.95), an accuracy of 85%, a sensitivity of 86%, a specificity of 86%, and an F1-score of 0.74. In addition, the internal representation features learned by the model showed clear differentiation across molecular subtype groups. Such a deep learning system can automatically predict triple-negative breast cancer preoperatively and accurately. It may help to get to more precise and comprehensive management.Graphical

Abstract Background Whole-genome sequencing has gained interest for assessing antimicrobial resistance (AMR). However, multiple studies have shown a poor correlation between the β-lactam/β-lactamase inhibitor (BL/BLI) phenotype and β-lactamase gene presence/absence. In a recent publication, we found that amplification of β-lactamase encoding genes potentially contribute to BL/BLI resistance and here we explore if gene amplifications allow for better BL/BLI genotype/phenotype correlations. Methods We selected 109 E. coli bacteremia isolates and performed ETests (bioMerieux, Inc) to determine AST profiles using the following BL/BLI antibiotics: ampicillin/sulbactam (SAM), amoxicillin/clavulanic acid (AMC), and piperacillin/tazobactam (TZP). CLSI M100 guidelines (2018) were used to stratify isolates into 4 groups: SAM/AMC/TZP susceptible (Group 1), SAM resistant only (Group 2), SAM/AMC resistant (Group 3), and SAM/AMC/TZP resistant (Group 4). Short-read whole genome sequencing was performed on these isolates and analyzed by identifying AMR genes, establishing multi-locus sequence type (MLST), estimating copy number variants (CNV), and investigating blaTEM-1 promoter regions in relation to phenotype. A t-Distributed Stochastic Neighbor Embedding (t-SNE) clustering method and a core-genome, maximum-likelihood phylogeny was generated to group isolates based on AMR gene content and CNV data. Results We found 34 different MLST groups with the three most common sequence types ST131 (n=41), ST1193 (n=12), and ST648 (n=6). Group 1 and Group 2 isolates had similar copy number estimates of β-lactamase blaTEM-1 (1.8X and 1.5X respectively) while there was an increase for both Group 3 (3.7X) and Group 4 (8.2X) isolates (Table 1). Additionally, blaOXA-1 was primarily present in Group 4 isolates (63%) with elevated copy numbers (5.2X). We did not observe associations of blaTEM-1 promoter regions with each of the BL/BLI Groups. Our t-SNE analysis shows that Group 1, Group 3, and Group 4 isolates cluster independently. Core Genome Maximum-likelihood Phylogeny. Midpoint-rooted core-genome maximum-likelihood phylogeny comparing core elements (MLST) to non core elements (CNV). AMR CNV and presence/absence associate better with BL/BLI phenotypes than do core elements. Conclusion There is a clear delineation between fully susceptible and fully resistant BL/BLI E. coli isolates when gene amplification and presence/absence of narrow-spectrum β-lactamases are considered, which should be considered for BL/BLI resistance prediction models. Disclosures David E. Greenberg, MD, Shionogi: Grant/Research Support.

Comparing the performance of classic voice-driven assistive systems for dysarthric speech

PurposeThe current study aims to test the hypothesis that the ratio of T1-weighted image to T2-weighted image signal intensity (T1w/T2w-Ratio: rT1/T2), an imaging surrogate developed for myelin integrity, is predictive of histologically lower-grade glioma's IDH mutation status.Materials and Methods25 histologically and molecularly confirmed lower-grade glioma patients with eight IDH-wildtype (IDHwt) and 17 IDH-mutant (IDHmt) tumors at Asahikawa Medical University Hospital (AMUH) were used as a test cohort. Twenty-nine patients (IDHwt: 13, IDHmt: 16) from Osaka International Cancer Institute (OICI) and 101 patients from the Cancer Imaging Archive (TCIA) / Cancer Genome Atlas (TCGA) dataset (IDHwt: 19, IDHmt: 82) were used as external cohorts. rT1/T2 images were calculated from T1- and T2-weighted images using a recommended signal correction. The relationship between the mean rT1/T2 (mrT1/T2) and the IDH mutation status was investigated. Moreover, t-Distributed Stochastic Neighbor Embedding (t-SNE) investigated the difference in MRI qualities and characteristics between the three cohorts.ResultsThe test cohort at AMUH revealed that mrT1/T2 of IDHwt tumors was significantly higher than that of IDHmt tumors (p < 0.05) and that the optimal cut-off of mrT1/T2 for discriminating IDHmt was 0.666-0.677, (AUC = 0.75, p < 0.05), which finding was validated by the external domestic cohort at OICI (AUC = 0.75, p = 0.02). However, the external international cohort deriving from TCIA/TCGA could not validate this (AUC = 0.63, p = 0.08). t-SNE analysis revealed that the difference in image characteristics within the cohort was more diverse for the TCIA/TCGA than for the two domestic cohorts.ConclusionThe current study revealed that mrT1/T2 was able to discriminate IDHwt and IDHmt tumors in two domestic cohorts significantly. This was not validated by the TCIA/TCGA cohort due to the wide variety in the original imaging characteristics of the TCIA/TCGA cohort.

Thymomas are malignant thymic epithelial tumors that are difficult to diagnose due to their rarity and complex diagnostic criteria. They represent a morphologically heterogeneous class of tumors mainly defined by "organo-typical" architectural features and cellular composition. The diagnosis of thymoma is burdened with a high level of inter-observer variability and the problem that some type-specific morphological alterations are more on the continuum than clear-cut. Methylation pattern-based classification may help to increase diagnostic precision, particularly in borderline cases. We applied array-based DNA methylation analysis to a set of 113 thymomas with stringent histological annotation. Unsupervised clustering and t-SNE analysis of DNA methylation data clearly segregated thymoma samples mainly according to the current WHO classification into A, AB, B1, B2, B2/B3, B3, and micronodular thymoma with lymphoid stroma. However, methylation analyses separated the histological subgroups AB and B2 into two methylation classes: mono-/bi-phasic AB-thymomas and conventional/"B1-like" B2-thymomas. Copy number variation analysis demonstrated methylation class-specific patterns of chromosomal alterations. Our study demonstrates that the current WHO classification is generally well reflected at the methylation level but suggests that B2- and AB-thymomas are (epi)genetically heterogeneous. Methylation-based classifications could help to refine diagnostic criteria for thymoma classification, improve reproducibility, and may affect treatment decisions.

Abstract BACKGROUND Spinal meningiomas account for 1.2-12 % of all meningiomas and 25-45 % of all spinal tumours. 20 % of intracranial, but only 4.6 % of spinal meningiomas recur requiring additional treatment. Whereas the classification of intracranial meningiomas has evolved considerably in recent years and uses genetic and epigenetic parameters, the classification of spinal meningiomas is based solely on histopathological findings. By embedding epi-/genetic features, the prognosis of intracranial meningiomas could be significantly improved, which is still lacking for spinal meningiomas. In our work, we integrated genetic and epigenetic parameters into the classification of spinal meningiomas. METHODS We performed epi-/genetic profiling of 50 spinal meningiomas. 497 intracranial meningiomas served as a reference cohort. Copy number variations (CNV) were inferred from the methylation data. Principal component (PCA) and t-SNE analysis were conducted. Clinical and histopathological parameters (location, size, recurrence, WHO°, pathological subtype) were correlated with methylation signatures using the DKFZ brain tumour classifier. RESULTS The methylation signature of spinal meningiomas matched to that of intracranial meningiomas (50/50), although meningioma subgroup assignment was achieved in only 13/50 cases. PCA and t-SNE analysis showed that most spinal meningiomas separate from cranial meningiomas and form two distinct clusters. Cluster 1 matched the methylation class ben-2, while cases in cluster 2 were heterogenous and had a low MSC score. Cases of cluster 1 were located in the upper spine, are more common in males and had an AKT1E17K mutation. NF2 mutations were found mainly in the second cluster, in line with a chr.22 q loss. Interestingly 4 tumors did not associate with the two spinal meningioma clusters and had a particular higher recurrence rate. CONCLUSION Genetic and epigenetic profiling of spinal meningiomas identifies two distinct classes of spinal meningiomas, which may allow an improved prognosis that could lead to a better guidance for adjuvant therapy.

Abstract INTRODUCTION: Plasma extracellular vesicles (EVs) have been shown as a promising source for biomarker identification in glioblastoma (GBM) and could help differential diagnosis, treatment evaluation and tumor progression monitoring. These EVs are enriched in molecular signatures indicative of their cell origins, giving an indication of the key players in this pathology. In this project, we aimed to identify diagnostic biomarkers for GBM plasma EVs and their cells of origin. METHODS: Plasma EV samples were prepared following the MIFlowCyt-EV guideline of the International Society for Extracellular Vesicles, then stained for EV markers (CD9/CD63/CD81) and markers indicative of cell origins (CD31/CD45/CD41a/CD11b). Actin phalloidin was used as a negative marker. Stained plasma samples were analyzed using a Cytek Aurora flow cytometer. Percentages of different EV subpopulations were analyzed and compared between GBM and normal donor (ND) plasma EVs (reference group). Further clustering analysis was performed on EV events by t-distributed stochastic neighbor embedding (t-SNE) and self-organizing maps on flow cytometry data (FlowSOM) analysis. The predictive value of multiparametric qualities derived from the reference group was tested in blinded test group samples. RESULTS: Percentages of CD9, CD81, and CD11b positive EVs were higher in GBM patient plasma, while ND plasma had more CD41a positive EVs. GBM plasma EVs had unique multiparametric signatures compared to ND plasma EVs based on t-SNE and FlowSOM analysis. Our analysis also identified 15 distinct EV subpopulations which differed in size and various surface marker expression levels. Eight of these subpopulations were enriched for GBM EVs, while three were enriched for ND EVs. Our method of multiparametric analysis demonstrates high sensitivity and specificity in predicting disease status in human samples. CONCLUSIONS: GBM plasma EVs have a unique surface marker expression profile and distinct EV subpopulations compared to ND plasma EVs. Multiparametric signatures show promise as potential diagnostic markers of GBM.