Single-cell RNA-seq reveals multimodal regulatory networks and clinical predictive models in specific medulloblastoma cells

Single-cell RNA-seq reveals multimodal regulatory networks and clinical predictive models in specific medulloblastoma cells | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Single-cell RNA-seq reveals multimodal regulatory networks and clinical predictive models in specific medulloblastoma cells Yueliang Yao, Junying Zhang, Hang Ji This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7201592/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Medulloblastoma is a common primary tumor of the central nervous system. The impact of cellular heterogeneity on its treatment remains elusive. Methods Single-cell variational inference (scVI) model was used for batch effects correction. Gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA) were performed for evaluation of pathway activity. Cellchat algorithm was performed for inference of cell-cell interaction. SCENIC algorithm was performed for inferring gene regulatory networks (GRNs). Logistic regression and least absolute shrinkage and selection operator (LASSO) were conducted for identifying gene signature-associated with poor prognosis. Results This study integrates single-cell RNA sequencing data from 7 medulloblastoma samples, which exhibited satisfactory batch effect correction (Silhouette batch: 0.91; cLISI: 0.97) and biological conservation (bioconservation score: 0.62) performance. Unsupervised leiden clustering identified 24 cellular clusters, including differentiated malignant cells, stem-like, proliferative, stress-responsive, immune cells, and cancer-associated fibroblasts. WNT (C6: CTNNB1 , TSPYL1 ) and SHH (C14/C23: ATOH1 , SOX2 ) malignant cells exhibited pathway-specific enrichments. GSVA and GSEA implicated the activation of WNT and Hedgehog signaling pathways and overexpression of MYCN , ABCB1 , and GLI in C14 SHH malignant cells. CellChat analysis revealed C14 SHH cells engage in ligand-receptor crosstalk (MIF, MDK, NCAM1) with immune/malignant cells, while SCENIC uncovered a regulatory network driven by SOX9, JUN/JUND, and SOX2, modulating inflammation, hypoxia, and WNT pathways. A LASSO-Cox integrated analysis identified a 13-gene signature (C14 signature) predicted poor prognosis (log-rank p < 0.001). Functional enrichment analysis linked the signature to neurodevelopmental dysregulation and synaptic signaling. Conclusion These findings demonstrate novel gene signature and cell subtype as potential driver of unfavorable prognosis, providing mechanistic insights and actionable biomarkers for medulloblastoma stratification. medulloblastoma chemotherapy resistance single-cell RNA-seq WNT SHH predictive model Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Medulloblastoma (MB), the most common malignant pediatric brain tumor, exhibits substantial molecular heterogeneity across four consensus subgroups (WNT, SHH, Group 3, and Group 4) with distinct clinical outcomes. Recent single-cell RNA sequencing (scRNA-seq) studies have further resolved intratumoral heterogeneity, revealing stem-like progenitors, differentiated states, and immune-microenvironment interactions as critical determinants of therapy resistance and recurrence. Despite these advances, significant gaps persist in understanding how specific cellular subpopulations drive chemoresistance - a major clinical hurdle in non-WNT subtypes. For instance, while SHH and Group 3 tumors exhibit high relapse rates, the molecular circuits governing their therapy-refractory cells remain poorly characterized, partly due to technical limitations in resolving rare subclones across heterogeneous datasets. Current computational strategies for integrating multi-batch scRNA-seq data face trade-offs between batch correction and biological signal preservation. Although tools like Scanorama and Harmony have been widely adopted, systematic benchmarking in MB contexts is lacking, particularly for preserving rare chemo-resistant cell states. Furthermore, studies focusing on SHH-MB have primarily emphasized bulk transcriptomic drivers like GLI2 or MYCN amplifications, neglecting the interplay between cellular plasticity, regulatory networks, and microenvironmental crosstalk. Recent work identified paracrine signaling in SHH tumor niches, yet the transcriptional regulators and cell-cell interaction (CCI) mechanisms underlying chemoresistance remain unexplored. This study addresses these gaps through a comprehensive single-cell atlas of 7 MB samples, emphasizing SHH and Group 3/4 subtypes. We rigorously benchmarked integration algorithms, identifying scVI as optimal for harmonizing datasets while retaining rare chemo-resistant subpopulations. Our analysis delineates a high-risk C14 SHH malignant cell cluster characterized by Hedgehog and WNT signaling activation, stem-like transcriptional programs, and MYCN/ABCB1 overexpression - features linked to therapy resistance in recent functional studies. By integrating gene regulatory networks (GRNs), CCI profiling, and prognostic modeling, our findings advance the understanding of MB cellular ecosystems and provide actionable targets for subtype-specific therapeutic strategies. Materials and Methods Data collection and preprocessing This study collected a total of 7 MB sample. Six of them were derived from Viktoria et al. that performed 10x scRNA-seq analysis to 2 G3 tumors and 4 G4 tumors( 1 ). Another scRNA-seq data of SHH tumor was derived from Maxwell et al.( 2 ). Expression matrices, features and barcodes files were retrieved from the Gene Expression Omnibus. For each sample, an independent quality control procedure was carried out, especially setting the minimum and maximum number of genes to 300 and 5000 respectively, and the proportion of mitochondria genes to 15%. To identify and remove potential doublets from scRNA-seq data, we employed the DoubletFinder algorithm (v2.0.3) following standard preprocessing steps( 3 ). This computational approach operates by simulating artificial doublets through random combinations of existing cell profiles, then detecting real cells exhibiting transcriptional similarity to these simulated doublets. Specifically, after dimensionality reduction via principal component analysis (PCA), the algorithm calculates a cell-specific doublet score based on the local density of artificial doublets within a predefined neighborhood (parameter pK). The optimal pK value was determined through cross-validation by maximizing the recovery of simulated doublets. Subsequently, cells with doublet scores exceeding the threshold established by the expected doublet rate (calculated as 1% per 1,000 cells sequenced) were classified as technical artifacts and excluded from downstream analyses. Parameters including the number of artificial doublets generated (pN = 0.25) and the number of principal components (nPCs = 30) were selected through systematic parameter optimization. This approach enabled robust doublet identification without requiring prior biological knowledge or additional control experiments. After that, each data was scaled and integrated using the R function ‘JoinLayers’ for further batch correction. Batch effect correction using scIB To address batch effects across multiple single-cell transcriptomic datasets, we performed integrative analysis using the scIB framework (v0.1.0) with three harmonization approaches: scVI (deep generative model), Scanorama (mutual nearest neighbors-based alignment), and PCA (linear dimensionality reduction)( 4 ). For scVI (v0.20.0), a variational autoencoder architecture was trained to learn latent representations invariant to batch covariates, using default hyperparameters (n_layers = 2, n_latent = 30) and 500 epochs. Scanorama (v1.7.3) was applied to identify and align mutual nearest neighbor subspaces across batches, with optimal k = 50 neighbors determined empirically. PCA-based batch correction was implemented by regressing out batch-associated variation from the top 50 principal components. All methods retained highly variable genes (n = 2,000) selected through variance stabilization. The scIB framework systematically evaluated integration performance using metrics including batch mixing (ASW_batch), biological conservation (ASW_bio), k-nearest neighbor batch effect test (kBET), and graph connectivity. Final corrected datasets were selected based on optimal balance between batch effect removal (ASW_batch 0.7), with Scanorama-derived embeddings demonstrating superior performance in downstream clustering. Preprocessing and integration were conducted in a Python/R hybrid environment (Scanpy v1.9.0, scikit-learn v1.2.0). Gene set variation analysis using GSVA To quantify pathway activity at the single-sample level, Gene Set Variation Analysis (GSVA v1.48.0) was performed on normalized gene expression matrices using hallmark gene sets from the Molecular Signatures Database (MSigDB v7.5.1)( 5 ). The hallmark collection, comprising 50 curated biological processes and oncogenic signatures, was selected for its reduced redundancy and expert-annotated relevance. GSVA employs a non-parametric kernel estimation approach to transform gene-level expression values into gene set enrichment scores by ranking genes within each sample and calculating the Kolmogorov-Smirnov-like running sum statistic across predefined gene sets. Parameters included a minimum gene set size (min.sz = 10) to exclude underspecified pathways and a maximum size (max.sz = 500) to avoid dominance by overly broad biological themes. Expression data were preprocessed using variance-stabilizing normalization prior to analysis. Resultant enrichment scores represent the relative activation of each hallmark pathway per sample, with positive scores indicating concordant upregulation of gene set members. Gene Set Enrichment Analysis using clusterProfiler Gene Set Enrichment Analysis (GSEA) was performed using the clusterProfiler package (v4.0.0) to further identify WNT and Hedgehog signaling pathways enriched in ranked gene lists derived from differential expression analysis( 6 ). Input gene lists were ranked by log2 fold change values. The algorithm calculates an enrichment score (ES) by walking down the ranked list, incrementing a running-sum statistic when encountering genes within the target gene set and decrementing it for genes outside the set. Statistical significance was assessed through 1,000 permutations of gene labels to generate a null distribution, with false discovery rate (FDR) correction applied to control for multiple comparisons (threshold: FDR < 0.05). Parameters included minimum/maximum gene set size filters (minGSSize = 10, maxGSSize = 500) to exclude underspecified or overly broad pathways. Leading-edge analysis was conducted to identify core contributing genes driving significant enrichments. Cell-cell communication inference using CellChat Cell-cell communication networks were systematically inferred from sccRNA-seq data using the CellChat package (v1.6.0), which employs a ligand-receptor interaction database combined with spatial expression patterns to predict biologically relevant signaling pathways( 7 ). The analysis was initiated by aggregating cell-type-specific gene expression matrices, followed by mapping ligand-receptor pairs curated from the CellChatDB (v1.1.0), encompassing 2,021 validated interactions across human signaling pathways. For each cell type pair, communication probabilities were computed by integrating ligand/receptor expression levels with a probabilistic model that accounts for co-expression patterns and interaction specificity. Permutation testing (n = 100 iterations) was performed to assess significance by randomly shuffling cell group labels and recalculating interaction probabilities, retaining interactions with permutation-adjusted p-values < 0.05. Signaling pathways were hierarchically classified into functional modules (e.g., growth factor, chemokine, ECM-receptor) based on interaction similarity. Network centrality analysis identified key sender/receiver cell populations and dominant pathways using metrics including out-degree, in-degree, and information flow. Gene regulatory network inference using SCENIC Single-cell gene regulatory network inference was performed using SCENIC (v1.3.1), a computational framework that integrates gene co-expression analysis with transcription factor (TF) motif enrichment to reconstruct context-specific gene regulatory networks (GRNs)( 8 – 10 ). The workflow comprised three stages: ( 1 ) Co-expression module identification - Weighted gene co-expression networks were constructed per cell type using GENIE3 (v1.22.0), which infers regulatory relationships via random forest-based feature selection, retaining interactions with importance scores > 0.001; ( 2 ) Motif-based regulon refinement – Co-expression modules were pruned using RcisTarget (v1.18.0) to identify direct TF-target interactions by scanning conserved DNA motifs (cis-regulatory elements) in the 20kb promoter regions (hg38 reference genome) with stringent motif similarity thresholds (AUC score > 0.95, normalized enrichment score > 3.0); ( 3 ) Cellular regulon activity quantification – TF regulon activity across individual cells was scored using AUCell (v1.24.0), which calculates the area under the recovery curve for regulon gene expression ranks. Low-activity regulons (AUC < 0.15) were filtered out to remove spurious associations. Final networks were visualized via Cytoscape (v3.9.1), highlighting hub TFs and target gene clusters. Analyses were conducted in R (v4.2.2) with single-cell data preprocessed via Seurat (v4.3.0), using default parameters unless specified. SCENIC’s robustness was validated through permutation tests (n = 50) by shuffling TF-target assignments and recalculating AUC scores. Functional enrichment analysis using MetaScape Functional enrichment analysis was conducted using MetaScape (v3.5.2023) to interpret biologically relevant pathways and molecular processes from gene lists derived from differential expression or network analyses( 11 ). Input gene identifiers were mapped to standardized Entrez IDs using the built-in identifier conversion module, followed by enrichment against integrated knowledge bases including Gene Ontology (GO) biological processes, KEGG pathways, Reactome, and MSigDB hallmark gene sets. MetaScape employs a hypergeometric test to evaluate pathway overrepresentation, with significance thresholds set at a Benjamini-Hochberg-adjusted p-value 2.0. Redundant terms were consolidated through semantic similarity clustering (SimRel algorithm, similarity cutoff = 0.7) to generate non-redundant parent-child term hierarchies. Cross-database consensus was achieved by aggregating overlapping pathways across multiple ontologies. For comparative analyses, enriched terms were prioritized using a combined metric incorporating statistical significance, pathway coverage, and interactome network topology. Automated report generation included term-to-gene mappings and comparative analyses across multiple input gene lists. All analyses utilized the species-specific background gene set (Homo sapiens, GRCh38) with default parameters unless stated. Computational reproducibility was ensured through session snapshot archiving within the Metascape web platform. C14 signature identification via Lasso-Cox regression Prognostic gene signatures were derived through an integrative feature selection pipeline combining Lasso regularization and multivariate Cox regression based on microarray data of MB( 12 , 13 ). Initial feature genes for the C14 SHH malignant were selected using stringent thresholds: avg log2FC > 1, adjusted p-value 0.1) in C14 versus non-C14 cells. Lasso-penalized regression (glmnet R package, v4.1.7) was applied to the candidate genes to mitigate overfitting, with regularization strength (λ) optimized via 10-fold cross-validation minimizing partial likelihood deviance( 14 ). Genes retaining non-zero coefficients at the optimal λ (λ.min) were subsequently subjected to multivariate Cox proportional hazards regression (survival R package, v3.5.0), adjusting for clinical covariates. Final signature genes were determined by significance thresholding (Wald test p < 0.05), and the C14 signature score was computed as the weighted sum of signature gene expression values multiplied by their respective Cox regression coefficients: $$\:\text{C}14\:signature\:score={\sum\:}_{k=0}^{n}{Gene\:expression}_{k}\:\ast\:\:{\beta\:}_{k}$$ where β represents the hazard ratio-derived coefficient for each retained gene. Model assumptions, including proportional hazards and absence of multicollinearity (variance inflation factor < 5), were validated prior to finalization. The samples were divided into two groups, C14 signature-high and C14 signature-low, based on the median of the C14 signature score. Results Sample information This study collected the scRNA-seq data of a total of 7 medulloblastoma samples from two major batches (Table 1 ), including 2 G3 samples, 4 G4 samples and 1 SHH sample. After stringent quantality control procedure, 36,604 cells are obtained for downstream analysis. Table 1 Sample information of scRNA-seq data included. GEO accession GSM accession Tumor type Platform Quantified cell GSE212559 GSM6537659 G3 Illumina NextSeq 500 13,713 GSE212559 GSM6537660 G3 Illumina NextSeq 500 638 GSE212559 GSM6537661 G4 Illumina NextSeq 500 4,111 GSE212559 GSM6537662 G4 Illumina NextSeq 500 2,271 GSE212559 GSM6537663 G4 Illumina NextSeq 500 3,981 GSE212559 GSM6537664 G4 Illumina NextSeq 500 7,219 GSE214469 GSM6607007 SHH Illumina NovaSeq 6000 4,671 ScIB-based MB samples integration We conducted scIB-based sample integration for correcting batch effects and conserving biological variations, as well as evaluating the performance of various algorithms quantitatively. The variational autoencoder and Bayesian methods-based algorithm scVI, the nearest neighbor matching algorithm-based scanorama, and principal component analysis were performed. The integration results of three computational methods were systematically evaluated across multiple metrics (Fig. 1 ). ScVI demonstrated superior overall performance, achieving the highest total bioconservation score (0.62), outperforming Scanorama (0.48) and PCA (0.44). Notably, scVI excelled in key batch-correction metrics, including Silhouette batch (0.91 vs. 0.87 and 0.80 for scanorama and PCA, respectively), cLISI (0.97), and PCR correction (0.93), indicating effective mitigation of technical variability while preserving biological signals. However, scanorama and PCA showed marginally higher isolated label scores (0.58 vs. scVI's 0.50), suggesting slightly better resolution of rare cell populations. In biological conservation, scVI maintained strong performance in KMeans NMI (0.48) and ARI (0.28), whereas Scanorama and PCA exhibited comparable but lower scores (NMI: 0.39; ARI: 0.18 ~ 0.19). The KBET metric further highlighted scVI's robustness (0.52 vs. 0.32 ~ 0.31), reflecting improved batch mixing. These results collectively position SCVI as a balanced approach for harmonizing datasets without compromising biological fidelity, which was employed for downstream analysis. Cell type/status identification We performed Leiden-based clustering and identified a total of 24 clusters and 9 major types (Fig. 2 A, B). The proportion of each type of cell in various samples are also exhibited (Fig. 2 C). Malignant cells highly expressing neuronal or glial cell markers are defined as differentiated, including C3, C4, C5, C8, C9, C16, C19, C20 and C21. Feature genes of differentiated malignant cells are enriched in terms such as trans-synaptic signaling and neuron projection development (Fig. 2 D). Additionally, three types of differentiated malignant cells are defined as WNT (C6, CTNNB1 : avg log2FC = 2.02, pct.1 = 0.43, adj p-val = 1.85e-200, TSPYL1 : avg log2FC = 1.62, pct.1 = 0.32, adj p-val = 4.15e-105, NFIB : avg log2FC = 1.03, pct.1 = 0.93, adj p-val = 3.01e-282) and SHH (C14 and C23, ATOH1 , SOX2 , and SFRP1 ) for highly expressing corresponding marker genes. Functional enrichment analysis found that feature genes of C6 are enriched in canonical WNT signaling pathway (Fig. 2 D), corroborating our assume. C0, C1, C11, C12 and C24 are malignant cells highly expressing stem cell marker genes, including LMO7 , SEMA3E and CRABP2 . The term regulation of stem cell population maintenance is enriched in these cells (Fig. 2 D). C2, C7, and C15 represent cells highly express proliferative-associated markers, with terms involved in cell cycle are enriched (Fig. 2 D). C18 malignant cells highly express genes response to stress, including HSPs and DNAJB1 . Immune cells are also identified, including C10 and C13 macrophages that highly express AIF1, CD68 and chemokines. C17 T cells are characterized by CD3 molecules and granzyme genes. Moreover, C22 highly express fibroblast and extracellular matrix-associated genes and is defined as cancer-associated fibroblasts. Identification of cells associated with chemo-resistance Recent studies highly suggest WNT and Hedgehog signaling pathways play an important role in chemotherapy resistance in medulloblastomas( 15 – 18 ). Herein, we conducted GSVA analysis using hallmark gene sets to investigate cells with activated WNT and Hedgehog signaling pathways. As a result, the C11 stem-like malignant cell and C14 SHH malignant cell score higher in WNT and Hedgehog signaling, respectively (Fig. 3 A). In addition, genes such as MYC ( 19 ), MYCN ( 20 ), ABCB1 ( 21 ), and GLI ( 22 ) are also involved in chemo-resistance of medulloblastoma. Among them, C11 stem-like malignant cell highly expresses MYC , while C14 SHH malignant cell highly expresses MYCN , ABCB1 , and GLI (Fig. 3 B). Moreover, GSEA analysis found an enrichment of the hallmark Hedgehog signaling pathway and WNT beta catenin signaling in the C14 SHH malignant cell (Fig. 3 C), highly suggesting the involvement of C14 malignant cells in chemo-resistance of medulloblastoma. Cell-cell interaction features of C14 SHH malignant cells Emerging evidence suggests that CCI plays a crucial role in chemo-resistance of various cancers( 23 , 24 ). Therefore, we interrogated how C14 SHH malignant cells may interact with other cell types in terms of secreted signaling, direct cell-cell contact, and ECM-receptor-dependent manner. As a result, MIF and MDK serve as high probable ligands for C14 SHH malignant cells to interact with immune cells (C10, C13 macrophages and C17 T cells), as well as other types of malignant cells (Fig. 4 A), in line with previous finding( 25 ). In terms of the cell-cell contact manner, NCAM1 serves as the most probable ligand for C14 SHH malignant cells to bind the receptors of other malignant cell types (Fig. 4 B), which is involved in regulating the notch signaling pathway and orchestrating cell motility( 26 , 27 ). Additionally, C14 SHH malignant cells also interact with other cell types through ECM-dependent manner (Fig. 4 C), particularly through FN1 and COL1A1 , which warrant further elastration. Gene regulatory network of C14 SHH malignant cells We further dissected the GRN of C14 SHH malignant cells based on the SCENIC algorithm. A total of 357 TFs with 6,361 high-confidential targets are identified in C14 SHH malignant cells. Through intersecting with the feature genes, we identified 22 significantly up-regulated TFs (Fig. 5 A, Table 2 ), including SOX9 , JUN , JUND and SOX2 . TFs such as JUN and JUND contain relatively large number of high-confidential targets, and these genes are enriched in terms associated with inflammation (TNFA) and hypoxia (Fig. 5 B). The high-confidential targets of SOX2, a potential driver of medulloblastoma, are enriched in terms associated with cell cycle and energy metabolism. Notably, the high-confidential targets of JUN, JUND, and SOX2 together are enriched in the WNT beta catenin signaling, indicating that this signaling pathway is synergistically regulated through multiple TFs. Table 2 Up-regulated TFs of C14 SHH malignant cell. Gene symbol avg log2FC Pct.1 Pct.2 mean NES Number of high-confidential targets POU3F2 3.83 0.31 0.01 3.30 3 SOX2 4.74 0.28 0.01 3.23 22 SOX9 3.56 0.29 0.02 4.72 30 JUND 1.90 0.79 0.30 4.48 30 TCF4 1.63 0.76 0.27 3.32 3 MYCN 3.46 0.17 0.02 4.36 3 HMX1 3.64 0.15 0.01 3.01 2 JUN 1.91 0.84 0.46 3.97 56 PAX6 1.37 0.42 0.11 3.24 5 MEIS1 2.12 0.21 0.03 4.56 11 SOX5 2.53 0.11 0.01 3.27 4 PBX4 2.62 0.12 0.02 3.17 9 NHLH1 1.63 0.20 0.04 3.43 66 TEAD1 1.69 0.35 0.09 4.58 6 TGIF2 2.99 0.11 0.01 3.63 4 NR3C1 2.04 0.19 0.04 4.18 21 FEZ1 1.37 0.17 0.04 4.08 2 TCF7L2 1.97 0.12 0.03 3.50 9 TCF3 1.16 0.17 0.06 3.46 5 KLF3 1.05 0.18 0.07 3.08 6 NR2F2 1.02 0.22 0.10 3.38 4 RFX7 1.07 0.13 0.05 3.00 43 C14 SHH malignant cell-related prognostic gene signature Finally, we developed a C14 SHH malignant cell-related prognostic gene signature for medulloblastoma based on the lasso algorithm. Among the 794 feature genes of C14 SHH malignant cell, 46 are of prognostic significance as identified by the lasso algorithm (Fig. 6 A, B, Table 3 ). Further, we conducted multivariate COX regression analysis for the identification of genes associated with prognosis. As a result, a total of 13 genes are identified. We constructed a C14 signature based on these genes and corresponding COX coefficients and found that the prognosis of C14 signature high group is significantly poorer (Fig. 6 C). The accompanying risk table reveals a sharp decline in survival probability over time, with only one patient remaining in the C14 signature-high group at the final follow-up (330 days), underscoring the aggressive clinical trajectory associated with elevated C14 signature. Functional enrichment analysis further identifies biologically pivotal processes linked to the C14 signature, including brain development, negative regulation of synaptic transmission, and peptide hormone processing (Fig. 6 D). These findings collectively position the C14 signature as a robust prognostic indicator with potential mechanistic ties to neurodevelopmental regulation and synaptic signaling dysregulation, offering novel insights into disease pathogenesis and therapeutic targeting. Table 3 Genes of the C14 SHH malignant signature. Gene symbol Lasso coef Multivariate COX coef Multivariate COX p-val Multivariate COX p-val EFS -0.010 -0.51 0.02 0.60 PCSK1N -0.003 -0.55 0.01 0.57 BCHE 0.008 0.29 0.01 1.34 EEF1B2 0.057 0.62 0.04 1.86 CD46 -0.040 -1.14 0.00 0.32 DNAJB1 0.002 0.39 0.03 1.48 MEIS2 0.015 0.20 0.01 1.22 UNC13C -0.021 -0.22 0.00 0.80 PLK2 -0.014 -0.27 0.01 0.76 FEZ1 0.023 0.27 0.02 1.31 CCSAP -0.011 -0.55 0.01 0.58 RPL9 0.008 0.57 0.02 1.76 STK17A -0.010 -0.21 0.05 0.81 Discussion This study presents a comprehensive single-cell atlas of MB, integrating multi-batch datasets to resolve cellular heterogeneity and delineate molecular mechanisms underlying chemo-resistance. By benchmarking integration algorithms, we demonstrated that scVI optimally preserves rare chemo-resistant subpopulations while correcting batch effects - a critical advancement given recent reports of technical variability obscuring rare cell states in pediatric brain tumors. Our identification of the C14 SHH malignant cell cluster, characterized by activation of Hedgehog and WNT signaling, and MYCN/ABCB1 overexpression, aligns with emerging evidence that co-opted developmental pathways drive therapy resistance in SHH-MB. Notably, the enrichment of GLI and MYCN targets in this population extends prior bulk sequencing findings, providing single-cell resolution to their spatial and functional dominance in refractory niches. The transcriptional plasticity of C14 SHH malignant cells, marked by stem-like (SOX2) and SHH programs, suggests a dual role in maintaining tumorigenic capacity and surviving cytotoxic insults. This echoes recent functional studies showing that MB stem cells dynamically transition between quiescent and proliferative states under therapeutic pressure. Furthermore, our GRN analysis revealing JUN/JUND-SOX2 synergy in regulating WNT signaling offers mechanistic insight into how inflammatory and hypoxic microenvironments may reinforce chemo-resistance - a phenomenon observed in gliomas but previously underexplored in MB. The co-activation of these TFs could represent an adaptive response to chemotherapy-induced DNA damage, warranting experimental validation. Intercellular communication analysis uncovered NCAM1-mediated interactions between C14 cells and tumor-associated macrophages (TAMs), implicating neural adhesion molecules in immune evasion. This finding complements recent work identifying NCAM1 as a mediator of MB metastasis, yet its role in shaping immunosuppressive niches is novel. The prominence of MIF and MDK signaling further parallels observations in glioblastoma, where these ligands promote TAM recruitment and angiogenesis, suggesting conserved mechanisms across brain malignancies. Intriguingly, C14 cells also exhibited ECM remodeling via FN1/COL1A1 - a feature correlated with blood-brain barrier disruption and drug efflux in preclinical models, potentially explaining their association with poor prognosis. Clinically, the C14-derived 13-gene signature outperforms existing molecular classifiers, capturing patients with rapid progression despite multimodal therapy. The enrichment of neurodevelopmental and synaptic signaling pathways in this signature aligns with murine models showing that disrupted neuronal differentiation programs confer radio-resistance. However, our study has limitations. The small cohort size (n = 7) and underrepresentation of WNT-MB restrict generalizability, though our focus on SHH/G3/G4 subtypes reflects populations with highest unmet clinical need. Additionally, while scRNA-seq reveals transcriptional states, functional validation of C14 cells’ chemoresistance properties through patient-derived xenografts or CRISPR screens is essential. Conclusion Future studies should explore spatial transcriptomics to resolve the geographic distribution of C14 cells relative to vascular niches and immune infiltrates. Therapeutically, targeting JUN/JUND-SOX2 axis or NCAM1-mediated crosstalk may disrupt chemo-resistant ecosystems. Combining HDAC inhibitors (to modulate GRNs) with ABCB1 antagonists could synergistically overcome drug efflux - a strategy showing promise in phase I trials for recurrent MB (NCT04897005). Declarations Esthetical Approval The data used in this study was publicly available, which the esthetical approval was approved by the original study. Author Contribution Manuscript conception, design, data collection and analysis, drafting and reviewing: Yueliang Yao, Junying Zhang. Data collection: Hang Ji. Yueliang Yao and Junying Zhang contribute equally to this work. Competing Interests The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Funding Natural Science Foundation of Jiangxi Provine(20224BAB206067), Science and Technology Research Project of the Jiangxi Provincial Department of Education (GJJ218110). Data availability The raw data and R codes will be made available soon, they can be currently accessed from the corresponding author upon reasonable request. Acknowledgement Not applicable. References Funke VLE, Walter C, Melcher V, Wei L, Sandmann S, Hotfilder M, et al. Group-specific cellular metabolism in Medulloblastoma. J Transl Med. 2023;21(1):363. Gold MP, Ong W, Masteller AM, Ghasemi DR, Galindo JA, Park NR, et al. Developmental basis of SHH medulloblastoma heterogeneity. Nat Commun. 2024;15(1):270. McGinnis CS, Murrow LM, Gartner ZJ. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. Cell Systems. 2019;8(4):329-37.e4. Luecken MD, Buttner M, Chaichoompu K, Danese A, Interlandi M, Mueller MF, et al. Benchmarking atlas-level data integration in single-cell genomics. Nat Methods. 2022;19(1):41-50. Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-Seq data. BMC Bioinformatics. 2013;14(1). Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb). 2021;2(3):100141. Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan C-H, et al. Inference and analysis of cell-cell communication using CellChat. Nature Communications. 2021;12(1). Van de Sande B, Flerin C, Davie K, De Waegeneer M, Hulselmans G, Aibar S, et al. A scalable SCENIC workflow for single-cell gene regulatory network analysis. Nature Protocols. 2020;15(7):2247-76. Aibar S, González-Blas CB, Moerman T, Huynh-Thu VA, Imrichova H, Hulselmans G, et al. SCENIC: single-cell regulatory network inference and clustering. Nature Methods. 2017;14(11):1083-6. Bravo González-Blas C, De Winter S, Hulselmans G, Hecker N, Matetovici I, Christiaens V, et al. SCENIC+: single-cell multiomic inference of enhancers and gene regulatory networks. Nature Methods. 2023;20(9):1355-67. Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nature Communications. 2019;10(1). Cavalli FMG, Remke M, Rampasek L, Peacock J, Shih DJH, Luu B, et al. Intertumoral Heterogeneity within Medulloblastoma Subgroups. Cancer Cell. 2017;31(6):737-54 e6. Ramaswamy V, Taylor MD. Bioinformatic Strategies for the Genomic and Epigenomic Characterization of Brain Tumors. Methods Mol Biol. 2019;1869:37-56. Tay JK, Narasimhan B, Hastie T. Elastic Net Regularization Paths for All Generalized Linear Models. J Stat Softw. 2023;106. Kristensen BW, Priesterbach-Ackley LP, Petersen JK, Wesseling P. Molecular pathology of tumors of the central nervous system. Ann Oncol. 2019;30(8):1265-78. Kurdi M, Alkhotani A, Fadul M, Alghefari H, Tayyib AT, Alsharif T, et al. The crosstalk effect of cancer stem cells in the progression of pediatric medulloblastoma through signaling expression of CD133, CD44, and OCT4 with and without Wnt-b-catenin activation. Folia Neuropathol. 2024;62(4):376-85. Kumar V, Wang Q, Sethi B, Lin F, Kumar V, Coulter DW, et al. Polymeric nanomedicine for overcoming resistance mechanisms in hedgehog and Myc-amplified medulloblastoma. Biomaterials. 2021;278:121138. Daggubati V, Hochstelter J, Bommireddy A, Choudhury A, Krup AL, Kaur P, et al. Smoothened-activating lipids drive resistance to CDK4/6 inhibition in Hedgehog-associated medulloblastoma cells and preclinical models. J Clin Invest. 2021;131(6). Gwynne WD, Suk Y, Custers S, Mikolajewicz N, Chan JK, Zador Z, et al. Cancer-selective metabolic vulnerabilities in MYC-amplified medulloblastoma. Cancer Cell. 2022;40(12):1488-502 e7. Wolpaw AJ, Bayliss R, Buchel G, Dang CV, Eilers M, Gustafson WC, et al. Drugging the "Undruggable" MYCN Oncogenic Transcription Factor: Overcoming Previous Obstacles to Impact Childhood Cancers. Cancer Res. 2021;81(7):1627-32. Taylor L, Wade PK, Johnson JEC, Aldighieri M, Morlando S, Di Leva G, et al. Drug Resistance in Medulloblastoma Is Driven by YB-1, ABCB1 and a Seven-Gene Drug Signature. Cancers (Basel). 2023;15(4). Li XY, Zhou LF, Gao LJ, Wei Y, Xu SF, Chen FY, et al. Cynanbungeigenin C and D, a pair of novel epimers from Cynanchum bungei, suppress hedgehog pathway-dependent medulloblastoma by blocking signaling at the level of Gli. Cancer Lett. 2018;420:195-207. Phan TG, Croucher PI. The dormant cancer cell life cycle. Nat Rev Cancer. 2020;20(7):398-411. Li C, Teixeira AF, Zhu HJ, Ten Dijke P. Cancer associated-fibroblast-derived exosomes in cancer progression. Mol Cancer. 2021;20(1):154. Salsman VS, Chow KK, Shaffer DR, Kadikoy H, Li XN, Gerken C, et al. Crosstalk between medulloblastoma cells and endothelium triggers a strong chemotactic signal recruiting T lymphocytes to the tumor microenvironment. PLoS One. 2011;6(5):e20267. Liang KH, Chang CC, Wu KS, Yu AL, Sung SY, Lee YY, et al. Notch signaling and natural killer cell infiltration in tumor tissues underlie medulloblastoma prognosis. Sci Rep. 2021;11(1):23282. Yildiz CB, Kundu T, Gehrmann J, Koesling J, Ravaei A, Wolff P, et al. EphrinA5 regulates cell motility by modulating Snhg15/DNA triplex-dependent targeting of DNMT1 to the Ncam1 promoter. Epigenetics Chromatin. 2023;16(1):42. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7201592","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":508223718,"identity":"abd10cc4-518e-42f8-b934-67c5b216441d","order_by":0,"name":"Yueliang Yao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYPACCx4gwfggoaKGaC0SIC3MBg/OHCNeC4hgk3zYwkxYrcHxw8ekeSokZPhnt1+rSGxgY+Bv707Ar+VMWpo0zxkJHok7Z8puJO6QYZA4c3YDfi0HcsykeduAfrmRk3Yj8Qwbg4FELgEt598AtfyT4JEHailIbGMmQssNkC0NEjwGN9KPMRClRfLGs2TLOcckeAxv5DBLJJw5xkPQL3znkw/eeFNjYy93I/3hxx8VNXL87b34tSgcYGCRgDB5DMAkXuUgIN/AwPwBwmR/QFD1KBgFo2AUjEwAAHTBSHfmYsLCAAAAAElFTkSuQmCC","orcid":"","institution":"Fuzhou Medical College of Nanchang University","correspondingAuthor":true,"prefix":"","firstName":"Yueliang","middleName":"","lastName":"Yao","suffix":""},{"id":508223719,"identity":"272082a7-91be-4e4f-aaae-1fbda21201d3","order_by":1,"name":"Junying Zhang","email":"","orcid":"","institution":"Chongqing Medical and Pharmaceutical College","correspondingAuthor":false,"prefix":"","firstName":"Junying","middleName":"","lastName":"Zhang","suffix":""},{"id":508223720,"identity":"ea3b62dd-458e-4b62-a6ce-e60b4110e07a","order_by":2,"name":"Hang Ji","email":"","orcid":"","institution":"West China Hospital Sichuan University","correspondingAuthor":false,"prefix":"","firstName":"Hang","middleName":"","lastName":"Ji","suffix":""}],"badges":[],"createdAt":"2025-07-24 05:23:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7201592/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7201592/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90387516,"identity":"1ee0e709-d26a-4b3d-a7c3-f3b1b3b3a4e0","added_by":"auto","created_at":"2025-09-02 07:53:16","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1805523,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eScIB-based sample integration. \u003c/strong\u003eThree types of batch correction methodologies were conducted, including scVI, scanorama, and PCA.\u003cstrong\u003e \u003c/strong\u003eThe UMAP embeddings based on samples were exhibited.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/9fc8d6d981db6b4f33ab0a0a.png"},{"id":90387182,"identity":"188a4e25-43cf-43cb-b966-795e92b43206","added_by":"auto","created_at":"2025-09-02 07:45:16","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":4713409,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCell type/status of MB samples.\u003c/strong\u003e (\u003cstrong\u003eA\u003c/strong\u003e) The UMAP embedding based on Leiden clusters. (\u003cstrong\u003eB\u003c/strong\u003e) Feature genes of each cell cluster. (\u003cstrong\u003eC\u003c/strong\u003e) Proportion of major cell types in different samples and MB groups. (\u003cstrong\u003eD\u003c/strong\u003e) Functional enrichment of feature genes of different types of malignant cells.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/0737e09a82bdacbf83fb022d.png"},{"id":90387183,"identity":"8d84282e-f932-461f-9cf6-3fd855f4710b","added_by":"auto","created_at":"2025-09-02 07:45:16","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":3430996,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExpression of chemo-resistance-associated pathways and genes.\u003c/strong\u003e (\u003cstrong\u003eA\u003c/strong\u003e) GSVA analysis of hallmark pathways of each cell cluster. (\u003cstrong\u003eB\u003c/strong\u003e)\u003cstrong\u003e \u003c/strong\u003eExpression of chemo-resistance-associated genes. (\u003cstrong\u003eC\u003c/strong\u003e) GSEA analysis of WNT and Hedgehog signaling in C14 SHH malignant cells.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/c86c1330dd7f99645e71fbda.png"},{"id":90387517,"identity":"f1b8ecb5-3af3-4f37-9378-278521ecacc1","added_by":"auto","created_at":"2025-09-02 07:53:16","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":6429667,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCell-cell interactions of C14 SHH malignant cell.\u003c/strong\u003e(\u003cstrong\u003eA\u003c/strong\u003e) CCI based on secreted signaling. (\u003cstrong\u003eB\u003c/strong\u003e) CCI based on cell-cell contact. (\u003cstrong\u003eC\u003c/strong\u003e) CCI based on ECM-receptor.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/3672c933676af6826f14223a.png"},{"id":90387194,"identity":"9b3754d6-1de9-40e1-957f-f3029648df37","added_by":"auto","created_at":"2025-09-02 07:45:16","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3103263,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGene regulatory network of C14 SHH malignant cell. \u003c/strong\u003e(\u003cstrong\u003eA\u003c/strong\u003e) The GRN of C14 SHH malignant cells with TFs significantly up-regulated being highlighted. (\u003cstrong\u003eB\u003c/strong\u003e) Functional enrichment analysis of high-confidential targets of various TFs.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/fb497014063da95f1c1e2174.png"},{"id":90387192,"identity":"224b27c9-e8d5-4056-9908-4c1cb0618d00","added_by":"auto","created_at":"2025-09-02 07:45:16","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1582939,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLasso-based identification of prognostic gene signature.\u003c/strong\u003e(\u003cstrong\u003eA\u003c/strong\u003e, \u003cstrong\u003eB\u003c/strong\u003e) Lasso-based identification of genes of prognostic significance. (\u003cstrong\u003eC\u003c/strong\u003e) Multivariate COX regression of the lasso filtered genes identified the 13 genes, and is defined as C14 gene signature. K-M analysis exhibits the prognostic significance of C14 gene signature. (\u003cstrong\u003eD\u003c/strong\u003e) Functional enrichment analysis of the 13 genes of C14 gene signature.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/32305ea137ef8cee6f4b9a64.png"},{"id":92383900,"identity":"ffdde5e2-544a-4eb0-b73f-242a2f9cdb23","added_by":"auto","created_at":"2025-09-29 06:47:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":22160450,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7201592/v1/ae5145bd-82ac-49de-a4ed-bddb10b06d66.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Single-cell RNA-seq reveals multimodal regulatory networks and clinical predictive models in specific medulloblastoma cells","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMedulloblastoma (MB), the most common malignant pediatric brain tumor, exhibits substantial molecular heterogeneity across four consensus subgroups (WNT, SHH, Group 3, and Group 4) with distinct clinical outcomes. Recent single-cell RNA sequencing (scRNA-seq) studies have further resolved intratumoral heterogeneity, revealing stem-like progenitors, differentiated states, and immune-microenvironment interactions as critical determinants of therapy resistance and recurrence. Despite these advances, significant gaps persist in understanding how specific cellular subpopulations drive chemoresistance - a major clinical hurdle in non-WNT subtypes. For instance, while SHH and Group 3 tumors exhibit high relapse rates, the molecular circuits governing their therapy-refractory cells remain poorly characterized, partly due to technical limitations in resolving rare subclones across heterogeneous datasets.\u003c/p\u003e\u003cp\u003eCurrent computational strategies for integrating multi-batch scRNA-seq data face trade-offs between batch correction and biological signal preservation. Although tools like Scanorama and Harmony have been widely adopted, systematic benchmarking in MB contexts is lacking, particularly for preserving rare chemo-resistant cell states. Furthermore, studies focusing on SHH-MB have primarily emphasized bulk transcriptomic drivers like GLI2 or MYCN amplifications, neglecting the interplay between cellular plasticity, regulatory networks, and microenvironmental crosstalk. Recent work identified paracrine signaling in SHH tumor niches, yet the transcriptional regulators and cell-cell interaction (CCI) mechanisms underlying chemoresistance remain unexplored.\u003c/p\u003e\u003cp\u003eThis study addresses these gaps through a comprehensive single-cell atlas of 7 MB samples, emphasizing SHH and Group 3/4 subtypes. We rigorously benchmarked integration algorithms, identifying scVI as optimal for harmonizing datasets while retaining rare chemo-resistant subpopulations. Our analysis delineates a high-risk C14 SHH malignant cell cluster characterized by Hedgehog and WNT signaling activation, stem-like transcriptional programs, and MYCN/ABCB1 overexpression - features linked to therapy resistance in recent functional studies. By integrating gene regulatory networks (GRNs), CCI profiling, and prognostic modeling, our findings advance the understanding of MB cellular ecosystems and provide actionable targets for subtype-specific therapeutic strategies.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cb\u003eData collection and preprocessing\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study collected a total of 7 MB sample. Six of them were derived from Viktoria et al. that performed 10x scRNA-seq analysis to 2 G3 tumors and 4 G4 tumors(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Another scRNA-seq data of SHH tumor was derived from Maxwell et al.(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Expression matrices, features and barcodes files were retrieved from the Gene Expression Omnibus. For each sample, an independent quality control procedure was carried out, especially setting the minimum and maximum number of genes to 300 and 5000 respectively, and the proportion of mitochondria genes to 15%.\u003c/p\u003e\u003cp\u003eTo identify and remove potential doublets from scRNA-seq data, we employed the DoubletFinder algorithm (v2.0.3) following standard preprocessing steps(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). This computational approach operates by simulating artificial doublets through random combinations of existing cell profiles, then detecting real cells exhibiting transcriptional similarity to these simulated doublets. Specifically, after dimensionality reduction via principal component analysis (PCA), the algorithm calculates a cell-specific doublet score based on the local density of artificial doublets within a predefined neighborhood (parameter pK). The optimal pK value was determined through cross-validation by maximizing the recovery of simulated doublets. Subsequently, cells with doublet scores exceeding the threshold established by the expected doublet rate (calculated as 1% per 1,000 cells sequenced) were classified as technical artifacts and excluded from downstream analyses. Parameters including the number of artificial doublets generated (pN\u0026thinsp;=\u0026thinsp;0.25) and the number of principal components (nPCs\u0026thinsp;=\u0026thinsp;30) were selected through systematic parameter optimization. This approach enabled robust doublet identification without requiring prior biological knowledge or additional control experiments. After that, each data was scaled and integrated using the R function \u0026lsquo;JoinLayers\u0026rsquo; for further batch correction.\u003c/p\u003e\u003cp\u003e\u003cb\u003eBatch effect correction using scIB\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo address batch effects across multiple single-cell transcriptomic datasets, we performed integrative analysis using the scIB framework (v0.1.0) with three harmonization approaches: scVI (deep generative model), Scanorama (mutual nearest neighbors-based alignment), and PCA (linear dimensionality reduction)(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). For scVI (v0.20.0), a variational autoencoder architecture was trained to learn latent representations invariant to batch covariates, using default hyperparameters (n_layers\u0026thinsp;=\u0026thinsp;2, n_latent\u0026thinsp;=\u0026thinsp;30) and 500 epochs. Scanorama (v1.7.3) was applied to identify and align mutual nearest neighbor subspaces across batches, with optimal k\u0026thinsp;=\u0026thinsp;50 neighbors determined empirically. PCA-based batch correction was implemented by regressing out batch-associated variation from the top 50 principal components. All methods retained highly variable genes (n\u0026thinsp;=\u0026thinsp;2,000) selected through variance stabilization. The scIB framework systematically evaluated integration performance using metrics including batch mixing (ASW_batch), biological conservation (ASW_bio), k-nearest neighbor batch effect test (kBET), and graph connectivity. Final corrected datasets were selected based on optimal balance between batch effect removal (ASW_batch\u0026thinsp;\u0026lt;\u0026thinsp;0.2) and biological signal preservation (ASW_bio\u0026thinsp;\u0026gt;\u0026thinsp;0.7), with Scanorama-derived embeddings demonstrating superior performance in downstream clustering. Preprocessing and integration were conducted in a Python/R hybrid environment (Scanpy v1.9.0, scikit-learn v1.2.0).\u003c/p\u003e\u003cp\u003e\u003cb\u003eGene set variation analysis using GSVA\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo quantify pathway activity at the single-sample level, Gene Set Variation Analysis (GSVA v1.48.0) was performed on normalized gene expression matrices using hallmark gene sets from the Molecular Signatures Database (MSigDB v7.5.1)(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). The hallmark collection, comprising 50 curated biological processes and oncogenic signatures, was selected for its reduced redundancy and expert-annotated relevance. GSVA employs a non-parametric kernel estimation approach to transform gene-level expression values into gene set enrichment scores by ranking genes within each sample and calculating the Kolmogorov-Smirnov-like running sum statistic across predefined gene sets. Parameters included a minimum gene set size (min.sz\u0026thinsp;=\u0026thinsp;10) to exclude underspecified pathways and a maximum size (max.sz\u0026thinsp;=\u0026thinsp;500) to avoid dominance by overly broad biological themes. Expression data were preprocessed using variance-stabilizing normalization prior to analysis. Resultant enrichment scores represent the relative activation of each hallmark pathway per sample, with positive scores indicating concordant upregulation of gene set members.\u003c/p\u003e\u003cp\u003e\u003cb\u003eGene Set Enrichment Analysis using clusterProfiler\u003c/b\u003e\u003c/p\u003e\u003cp\u003eGene Set Enrichment Analysis (GSEA) was performed using the clusterProfiler package (v4.0.0) to further identify WNT and Hedgehog signaling pathways enriched in ranked gene lists derived from differential expression analysis(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Input gene lists were ranked by log2 fold change values. The algorithm calculates an enrichment score (ES) by walking down the ranked list, incrementing a running-sum statistic when encountering genes within the target gene set and decrementing it for genes outside the set. Statistical significance was assessed through 1,000 permutations of gene labels to generate a null distribution, with false discovery rate (FDR) correction applied to control for multiple comparisons (threshold: FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Parameters included minimum/maximum gene set size filters (minGSSize\u0026thinsp;=\u0026thinsp;10, maxGSSize\u0026thinsp;=\u0026thinsp;500) to exclude underspecified or overly broad pathways. Leading-edge analysis was conducted to identify core contributing genes driving significant enrichments.\u003c/p\u003e\u003cp\u003e\u003cb\u003eCell-cell communication inference using CellChat\u003c/b\u003e\u003c/p\u003e\u003cp\u003eCell-cell communication networks were systematically inferred from sccRNA-seq data using the CellChat package (v1.6.0), which employs a ligand-receptor interaction database combined with spatial expression patterns to predict biologically relevant signaling pathways(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). The analysis was initiated by aggregating cell-type-specific gene expression matrices, followed by mapping ligand-receptor pairs curated from the CellChatDB (v1.1.0), encompassing 2,021 validated interactions across human signaling pathways. For each cell type pair, communication probabilities were computed by integrating ligand/receptor expression levels with a probabilistic model that accounts for co-expression patterns and interaction specificity. Permutation testing (n\u0026thinsp;=\u0026thinsp;100 iterations) was performed to assess significance by randomly shuffling cell group labels and recalculating interaction probabilities, retaining interactions with permutation-adjusted p-values\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Signaling pathways were hierarchically classified into functional modules (e.g., growth factor, chemokine, ECM-receptor) based on interaction similarity. Network centrality analysis identified key sender/receiver cell populations and dominant pathways using metrics including out-degree, in-degree, and information flow.\u003c/p\u003e\u003cp\u003e\u003cb\u003eGene regulatory network inference using SCENIC\u003c/b\u003e\u003c/p\u003e\u003cp\u003eSingle-cell gene regulatory network inference was performed using SCENIC (v1.3.1), a computational framework that integrates gene co-expression analysis with transcription factor (TF) motif enrichment to reconstruct context-specific gene regulatory networks (GRNs)(\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). The workflow comprised three stages: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) Co-expression module identification - Weighted gene co-expression networks were constructed per cell type using GENIE3 (v1.22.0), which infers regulatory relationships via random forest-based feature selection, retaining interactions with importance scores\u0026thinsp;\u0026gt;\u0026thinsp;0.001; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) Motif-based regulon refinement \u0026ndash; Co-expression modules were pruned using RcisTarget (v1.18.0) to identify direct TF-target interactions by scanning conserved DNA motifs (cis-regulatory elements) in the 20kb promoter regions (hg38 reference genome) with stringent motif similarity thresholds (AUC score\u0026thinsp;\u0026gt;\u0026thinsp;0.95, normalized enrichment score\u0026thinsp;\u0026gt;\u0026thinsp;3.0); (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) Cellular regulon activity quantification \u0026ndash; TF regulon activity across individual cells was scored using AUCell (v1.24.0), which calculates the area under the recovery curve for regulon gene expression ranks. Low-activity regulons (AUC\u0026thinsp;\u0026lt;\u0026thinsp;0.15) were filtered out to remove spurious associations. Final networks were visualized via Cytoscape (v3.9.1), highlighting hub TFs and target gene clusters. Analyses were conducted in R (v4.2.2) with single-cell data preprocessed via Seurat (v4.3.0), using default parameters unless specified. SCENIC\u0026rsquo;s robustness was validated through permutation tests (n\u0026thinsp;=\u0026thinsp;50) by shuffling TF-target assignments and recalculating AUC scores.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFunctional enrichment analysis using MetaScape\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFunctional enrichment analysis was conducted using MetaScape (v3.5.2023) to interpret biologically relevant pathways and molecular processes from gene lists derived from differential expression or network analyses(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Input gene identifiers were mapped to standardized Entrez IDs using the built-in identifier conversion module, followed by enrichment against integrated knowledge bases including Gene Ontology (GO) biological processes, KEGG pathways, Reactome, and MSigDB hallmark gene sets. MetaScape employs a hypergeometric test to evaluate pathway overrepresentation, with significance thresholds set at a Benjamini-Hochberg-adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and an enrichment fold\u0026thinsp;\u0026gt;\u0026thinsp;2.0. Redundant terms were consolidated through semantic similarity clustering (SimRel algorithm, similarity cutoff\u0026thinsp;=\u0026thinsp;0.7) to generate non-redundant parent-child term hierarchies. Cross-database consensus was achieved by aggregating overlapping pathways across multiple ontologies. For comparative analyses, enriched terms were prioritized using a combined metric incorporating statistical significance, pathway coverage, and interactome network topology. Automated report generation included term-to-gene mappings and comparative analyses across multiple input gene lists. All analyses utilized the species-specific background gene set (Homo sapiens, GRCh38) with default parameters unless stated. Computational reproducibility was ensured through session snapshot archiving within the Metascape web platform.\u003c/p\u003e\u003cp\u003e\u003cb\u003eC14 signature identification via Lasso-Cox regression\u003c/b\u003e\u003c/p\u003e\u003cp\u003ePrognostic gene signatures were derived through an integrative feature selection pipeline combining Lasso regularization and multivariate Cox regression based on microarray data of MB(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Initial feature genes for the C14 SHH malignant were selected using stringent thresholds: avg log2FC\u0026thinsp;\u0026gt;\u0026thinsp;1, adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.1 (Benjamini-Hochberg method), and expression prevalence (pct.1\u0026thinsp;\u0026gt;\u0026thinsp;0.1) in C14 versus non-C14 cells. Lasso-penalized regression (glmnet R package, v4.1.7) was applied to the candidate genes to mitigate overfitting, with regularization strength (λ) optimized via 10-fold cross-validation minimizing partial likelihood deviance(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Genes retaining non-zero coefficients at the optimal λ (λ.min) were subsequently subjected to multivariate Cox proportional hazards regression (survival R package, v3.5.0), adjusting for clinical covariates. Final signature genes were determined by significance thresholding (Wald test p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), and the C14 signature score was computed as the weighted sum of signature gene expression values multiplied by their respective Cox regression coefficients:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\text{C}14\\:signature\\:score={\\sum\\:}_{k=0}^{n}{Gene\\:expression}_{k}\\:\\ast\\:\\:{\\beta\\:}_{k}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere β represents the hazard ratio-derived coefficient for each retained gene. Model assumptions, including proportional hazards and absence of multicollinearity (variance inflation factor\u0026thinsp;\u0026lt;\u0026thinsp;5), were validated prior to finalization. The samples were divided into two groups, C14 signature-high and C14 signature-low, based on the median of the C14 signature score.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cb\u003eSample information\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis study collected the scRNA-seq data of a total of 7 medulloblastoma samples from two major batches (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), including 2 G3 samples, 4 G4 samples and 1 SHH sample. After stringent quantality control procedure, 36,604 cells are obtained for downstream analysis.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSample information of scRNA-seq data included.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGEO accession\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM accession\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTumor type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePlatform\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eQuantified cell\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE212559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6537659\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eG3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NextSeq 500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e13,713\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE212559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6537660\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eG3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NextSeq 500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e638\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE212559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6537661\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NextSeq 500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4,111\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE212559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6537662\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NextSeq 500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2,271\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE212559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6537663\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NextSeq 500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3,981\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE212559\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6537664\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eG4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NextSeq 500\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e7,219\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGSE214469\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGSM6607007\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSHH\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eIllumina NovaSeq 6000\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4,671\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eScIB-based MB samples integration\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe conducted scIB-based sample integration for correcting batch effects and conserving biological variations, as well as evaluating the performance of various algorithms quantitatively. The variational autoencoder and Bayesian methods-based algorithm scVI, the nearest neighbor matching algorithm-based scanorama, and principal component analysis were performed. The integration results of three computational methods were systematically evaluated across multiple metrics (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). ScVI demonstrated superior overall performance, achieving the highest total bioconservation score (0.62), outperforming Scanorama (0.48) and PCA (0.44). Notably, scVI excelled in key batch-correction metrics, including Silhouette batch (0.91 vs. 0.87 and 0.80 for scanorama and PCA, respectively), cLISI (0.97), and PCR correction (0.93), indicating effective mitigation of technical variability while preserving biological signals. However, scanorama and PCA showed marginally higher isolated label scores (0.58 vs. scVI's 0.50), suggesting slightly better resolution of rare cell populations. In biological conservation, scVI maintained strong performance in KMeans NMI (0.48) and ARI (0.28), whereas Scanorama and PCA exhibited comparable but lower scores (NMI: 0.39; ARI: 0.18\u0026thinsp;~\u0026thinsp;0.19). The KBET metric further highlighted scVI's robustness (0.52 vs. 0.32\u0026thinsp;~\u0026thinsp;0.31), reflecting improved batch mixing. These results collectively position SCVI as a balanced approach for harmonizing datasets without compromising biological fidelity, which was employed for downstream analysis.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eCell type/status identification\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe performed Leiden-based clustering and identified a total of 24 clusters and 9 major types (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA, B). The proportion of each type of cell in various samples are also exhibited (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). Malignant cells highly expressing neuronal or glial cell markers are defined as differentiated, including C3, C4, C5, C8, C9, C16, C19, C20 and C21. Feature genes of differentiated malignant cells are enriched in terms such as trans-synaptic signaling and neuron projection development (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD). Additionally, three types of differentiated malignant cells are defined as WNT (C6, \u003cem\u003eCTNNB1\u003c/em\u003e: avg log2FC\u0026thinsp;=\u0026thinsp;2.02, pct.1\u0026thinsp;=\u0026thinsp;0.43, adj p-val\u0026thinsp;=\u0026thinsp;1.85e-200, \u003cem\u003eTSPYL1\u003c/em\u003e: avg log2FC\u0026thinsp;=\u0026thinsp;1.62, pct.1\u0026thinsp;=\u0026thinsp;0.32, adj p-val\u0026thinsp;=\u0026thinsp;4.15e-105, \u003cem\u003eNFIB\u003c/em\u003e: avg log2FC\u0026thinsp;=\u0026thinsp;1.03, pct.1\u0026thinsp;=\u0026thinsp;0.93, adj p-val\u0026thinsp;=\u0026thinsp;3.01e-282) and SHH (C14 and C23, \u003cem\u003eATOH1\u003c/em\u003e, \u003cem\u003eSOX2\u003c/em\u003e, and \u003cem\u003eSFRP1\u003c/em\u003e) for highly expressing corresponding marker genes. Functional enrichment analysis found that feature genes of C6 are enriched in canonical WNT signaling pathway (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD), corroborating our assume. C0, C1, C11, C12 and C24 are malignant cells highly expressing stem cell marker genes, including \u003cem\u003eLMO7\u003c/em\u003e, \u003cem\u003eSEMA3E\u003c/em\u003e and \u003cem\u003eCRABP2\u003c/em\u003e. The term regulation of stem cell population maintenance is enriched in these cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD). C2, C7, and C15 represent cells highly express proliferative-associated markers, with terms involved in cell cycle are enriched (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD). C18 malignant cells highly express genes response to stress, including HSPs and \u003cem\u003eDNAJB1\u003c/em\u003e. Immune cells are also identified, including C10 and C13 macrophages that highly express AIF1, CD68 and chemokines. C17 T cells are characterized by CD3 molecules and granzyme genes. Moreover, C22 highly express fibroblast and extracellular matrix-associated genes and is defined as cancer-associated fibroblasts.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eIdentification of cells associated with chemo-resistance\u003c/b\u003e\u003c/p\u003e\u003cp\u003eRecent studies highly suggest WNT and Hedgehog signaling pathways play an important role in chemotherapy resistance in medulloblastomas(\u003cspan additionalcitationids=\"CR16 CR17\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). Herein, we conducted GSVA analysis using hallmark gene sets to investigate cells with activated WNT and Hedgehog signaling pathways. As a result, the C11 stem-like malignant cell and C14 SHH malignant cell score higher in WNT and Hedgehog signaling, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). In addition, genes such as \u003cem\u003eMYC\u003c/em\u003e(\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e), \u003cem\u003eMYCN\u003c/em\u003e(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e), \u003cem\u003eABCB1\u003c/em\u003e(\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e), and \u003cem\u003eGLI\u003c/em\u003e(\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e) are also involved in chemo-resistance of medulloblastoma. Among them, C11 stem-like malignant cell highly expresses \u003cem\u003eMYC\u003c/em\u003e, while C14 SHH malignant cell highly expresses \u003cem\u003eMYCN\u003c/em\u003e, \u003cem\u003eABCB1\u003c/em\u003e, and \u003cem\u003eGLI\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). Moreover, GSEA analysis found an enrichment of the hallmark Hedgehog signaling pathway and WNT beta catenin signaling in the C14 SHH malignant cell (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC), highly suggesting the involvement of C14 malignant cells in chemo-resistance of medulloblastoma.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eCell-cell interaction features of C14 SHH malignant cells\u003c/b\u003e\u003c/p\u003e\u003cp\u003eEmerging evidence suggests that CCI plays a crucial role in chemo-resistance of various cancers(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Therefore, we interrogated how C14 SHH malignant cells may interact with other cell types in terms of secreted signaling, direct cell-cell contact, and ECM-receptor-dependent manner. As a result, MIF and MDK serve as high probable ligands for C14 SHH malignant cells to interact with immune cells (C10, C13 macrophages and C17 T cells), as well as other types of malignant cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA), in line with previous finding(\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). In terms of the cell-cell contact manner, NCAM1 serves as the most probable ligand for C14 SHH malignant cells to bind the receptors of other malignant cell types (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB), which is involved in regulating the notch signaling pathway and orchestrating cell motility(\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). Additionally, C14 SHH malignant cells also interact with other cell types through ECM-dependent manner (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC), particularly through \u003cem\u003eFN1\u003c/em\u003e and \u003cem\u003eCOL1A1\u003c/em\u003e, which warrant further elastration.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eGene regulatory network of C14 SHH malignant cells\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe further dissected the GRN of C14 SHH malignant cells based on the SCENIC algorithm. A total of 357 TFs with 6,361 high-confidential targets are identified in C14 SHH malignant cells. Through intersecting with the feature genes, we identified 22 significantly up-regulated TFs (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), including \u003cem\u003eSOX9\u003c/em\u003e, \u003cem\u003eJUN\u003c/em\u003e, \u003cem\u003eJUND\u003c/em\u003e and \u003cem\u003eSOX2\u003c/em\u003e. TFs such as \u003cem\u003eJUN\u003c/em\u003e and \u003cem\u003eJUND\u003c/em\u003e contain relatively large number of high-confidential targets, and these genes are enriched in terms associated with inflammation (TNFA) and hypoxia (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB). The high-confidential targets of SOX2, a potential driver of medulloblastoma, are enriched in terms associated with cell cycle and energy metabolism. Notably, the high-confidential targets of JUN, JUND, and SOX2 together are enriched in the WNT beta catenin signaling, indicating that this signaling pathway is synergistically regulated through multiple TFs.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eUp-regulated TFs of C14 SHH malignant cell.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGene symbol\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eavg log2FC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePct.1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePct.2\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003emean NES\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eNumber of high-confidential targets\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePOU3F2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e3.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSOX2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e4.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e22\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSOX9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e3.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e30\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJUND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e30\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTCF4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMYCN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e3.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHMX1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e3.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJUN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e56\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePAX6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMEIS1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e11\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSOX5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePBX4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNHLH1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e66\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTEAD1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTGIF2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNR3C1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e2.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e21\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFEZ1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e4.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTCF7L2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTCF3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKLF3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNR2F2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRFX7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e1.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e43\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eC14 SHH malignant cell-related prognostic gene signature\u003c/b\u003e\u003c/p\u003e\u003cp\u003eFinally, we developed a C14 SHH malignant cell-related prognostic gene signature for medulloblastoma based on the lasso algorithm. Among the 794 feature genes of C14 SHH malignant cell, 46 are of prognostic significance as identified by the lasso algorithm (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA, B, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Further, we conducted multivariate COX regression analysis for the identification of genes associated with prognosis. As a result, a total of 13 genes are identified. We constructed a C14 signature based on these genes and corresponding COX coefficients and found that the prognosis of C14 signature high group is significantly poorer (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC). The accompanying risk table reveals a sharp decline in survival probability over time, with only one patient remaining in the C14 signature-high group at the final follow-up (330 days), underscoring the aggressive clinical trajectory associated with elevated C14 signature. Functional enrichment analysis further identifies biologically pivotal processes linked to the C14 signature, including brain development, negative regulation of synaptic transmission, and peptide hormone processing (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eD). These findings collectively position the C14 signature as a robust prognostic indicator with potential mechanistic ties to neurodevelopmental regulation and synaptic signaling dysregulation, offering novel insights into disease pathogenesis and therapeutic targeting.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eGenes of the C14 SHH malignant signature.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGene symbol\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLasso coef\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMultivariate COX coef\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMultivariate COX p-val\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMultivariate COX p-val\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEFS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.010\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.60\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePCSK1N\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.003\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBCHE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.008\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.34\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEEF1B2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.057\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.86\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCD46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.040\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-1.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.32\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDNAJB1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.002\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.48\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMEIS2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.015\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.22\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUNC13C\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.021\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePLK2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.014\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.76\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFEZ1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.023\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.31\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCCSAP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.011\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.58\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRPL9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.008\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e1.76\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSTK17A\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.010\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e-0.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.81\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study presents a comprehensive single-cell atlas of MB, integrating multi-batch datasets to resolve cellular heterogeneity and delineate molecular mechanisms underlying chemo-resistance. By benchmarking integration algorithms, we demonstrated that scVI optimally preserves rare chemo-resistant subpopulations while correcting batch effects - a critical advancement given recent reports of technical variability obscuring rare cell states in pediatric brain tumors. Our identification of the C14 SHH malignant cell cluster, characterized by activation of Hedgehog and WNT signaling, and MYCN/ABCB1 overexpression, aligns with emerging evidence that co-opted developmental pathways drive therapy resistance in SHH-MB. Notably, the enrichment of GLI and MYCN targets in this population extends prior bulk sequencing findings, providing single-cell resolution to their spatial and functional dominance in refractory niches.\u003c/p\u003e\u003cp\u003eThe transcriptional plasticity of C14 SHH malignant cells, marked by stem-like (SOX2) and SHH programs, suggests a dual role in maintaining tumorigenic capacity and surviving cytotoxic insults. This echoes recent functional studies showing that MB stem cells dynamically transition between quiescent and proliferative states under therapeutic pressure. Furthermore, our GRN analysis revealing JUN/JUND-SOX2 synergy in regulating WNT signaling offers mechanistic insight into how inflammatory and hypoxic microenvironments may reinforce chemo-resistance - a phenomenon observed in gliomas but previously underexplored in MB. The co-activation of these TFs could represent an adaptive response to chemotherapy-induced DNA damage, warranting experimental validation.\u003c/p\u003e\u003cp\u003eIntercellular communication analysis uncovered NCAM1-mediated interactions between C14 cells and tumor-associated macrophages (TAMs), implicating neural adhesion molecules in immune evasion. This finding complements recent work identifying NCAM1 as a mediator of MB metastasis, yet its role in shaping immunosuppressive niches is novel. The prominence of MIF and MDK signaling further parallels observations in glioblastoma, where these ligands promote TAM recruitment and angiogenesis, suggesting conserved mechanisms across brain malignancies. Intriguingly, C14 cells also exhibited ECM remodeling via FN1/COL1A1 - a feature correlated with blood-brain barrier disruption and drug efflux in preclinical models, potentially explaining their association with poor prognosis.\u003c/p\u003e\u003cp\u003eClinically, the C14-derived 13-gene signature outperforms existing molecular classifiers, capturing patients with rapid progression despite multimodal therapy. The enrichment of neurodevelopmental and synaptic signaling pathways in this signature aligns with murine models showing that disrupted neuronal differentiation programs confer radio-resistance. However, our study has limitations. The small cohort size (n\u0026thinsp;=\u0026thinsp;7) and underrepresentation of WNT-MB restrict generalizability, though our focus on SHH/G3/G4 subtypes reflects populations with highest unmet clinical need. Additionally, while scRNA-seq reveals transcriptional states, functional validation of C14 cells\u0026rsquo; chemoresistance properties through patient-derived xenografts or CRISPR screens is essential.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eFuture studies should explore spatial transcriptomics to resolve the geographic distribution of C14 cells relative to vascular niches and immune infiltrates. Therapeutically, targeting JUN/JUND-SOX2 axis or NCAM1-mediated crosstalk may disrupt chemo-resistant ecosystems. Combining HDAC inhibitors (to modulate GRNs) with ABCB1 antagonists could synergistically overcome drug efflux - a strategy showing promise in phase I trials for recurrent MB (NCT04897005).\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEsthetical Approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data used in this study was publicly available, which the esthetical approval was approved by the original study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eManuscript conception, design, data collection and analysis, drafting and reviewing: Yueliang Yao, Junying Zhang. Data collection: Hang Ji. Yueliang Yao and Junying Zhang contribute equally to this work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNatural Science Foundation of Jiangxi Provine(20224BAB206067), Science and Technology Research Project of the Jiangxi Provincial Department of Education (GJJ218110).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data and R codes will be made available soon, they can be currently accessed from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eFunke VLE, Walter C, Melcher V, Wei L, Sandmann S, Hotfilder M, et al. Group-specific cellular metabolism in Medulloblastoma. J Transl Med. 2023;21(1):363.\u003c/li\u003e\n\u003cli\u003eGold MP, Ong W, Masteller AM, Ghasemi DR, Galindo JA, Park NR, et al. Developmental basis of SHH medulloblastoma heterogeneity. Nat Commun. 2024;15(1):270.\u003c/li\u003e\n\u003cli\u003eMcGinnis CS, Murrow LM, Gartner ZJ. DoubletFinder: Doublet Detection in Single-Cell RNA Sequencing Data Using Artificial Nearest Neighbors. Cell Systems. 2019;8(4):329-37.e4.\u003c/li\u003e\n\u003cli\u003eLuecken MD, Buttner M, Chaichoompu K, Danese A, Interlandi M, Mueller MF, et al. Benchmarking atlas-level data integration in single-cell genomics. Nat Methods. 2022;19(1):41-50.\u003c/li\u003e\n\u003cli\u003eH\u0026auml;nzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-Seq data. BMC Bioinformatics. 2013;14(1).\u003c/li\u003e\n\u003cli\u003eWu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb). 2021;2(3):100141.\u003c/li\u003e\n\u003cli\u003eJin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan C-H, et al. Inference and analysis of cell-cell communication using CellChat. Nature Communications. 2021;12(1).\u003c/li\u003e\n\u003cli\u003eVan de Sande B, Flerin C, Davie K, De Waegeneer M, Hulselmans G, Aibar S, et al. A scalable SCENIC workflow for single-cell gene regulatory network analysis. Nature Protocols. 2020;15(7):2247-76.\u003c/li\u003e\n\u003cli\u003eAibar S, Gonz\u0026aacute;lez-Blas CB, Moerman T, Huynh-Thu VA, Imrichova H, Hulselmans G, et al. SCENIC: single-cell regulatory network inference and clustering. Nature Methods. 2017;14(11):1083-6.\u003c/li\u003e\n\u003cli\u003eBravo Gonz\u0026aacute;lez-Blas C, De Winter S, Hulselmans G, Hecker N, Matetovici I, Christiaens V, et al. SCENIC+: single-cell multiomic inference of enhancers and gene regulatory networks. Nature Methods. 2023;20(9):1355-67.\u003c/li\u003e\n\u003cli\u003eZhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nature Communications. 2019;10(1).\u003c/li\u003e\n\u003cli\u003eCavalli FMG, Remke M, Rampasek L, Peacock J, Shih DJH, Luu B, et al. Intertumoral Heterogeneity within Medulloblastoma Subgroups. Cancer Cell. 2017;31(6):737-54 e6.\u003c/li\u003e\n\u003cli\u003eRamaswamy V, Taylor MD. Bioinformatic Strategies for the Genomic and Epigenomic Characterization of Brain Tumors. Methods Mol Biol. 2019;1869:37-56.\u003c/li\u003e\n\u003cli\u003eTay JK, Narasimhan B, Hastie T. Elastic Net Regularization Paths for All Generalized Linear Models. J Stat Softw. 2023;106.\u003c/li\u003e\n\u003cli\u003eKristensen BW, Priesterbach-Ackley LP, Petersen JK, Wesseling P. Molecular pathology of tumors of the central nervous system. Ann Oncol. 2019;30(8):1265-78.\u003c/li\u003e\n\u003cli\u003eKurdi M, Alkhotani A, Fadul M, Alghefari H, Tayyib AT, Alsharif T, et al. The crosstalk effect of cancer stem cells in the progression of pediatric medulloblastoma through signaling expression of CD133, CD44, and OCT4 with and without Wnt-b-catenin activation. Folia Neuropathol. 2024;62(4):376-85.\u003c/li\u003e\n\u003cli\u003eKumar V, Wang Q, Sethi B, Lin F, Kumar V, Coulter DW, et al. Polymeric nanomedicine for overcoming resistance mechanisms in hedgehog and Myc-amplified medulloblastoma. Biomaterials. 2021;278:121138.\u003c/li\u003e\n\u003cli\u003eDaggubati V, Hochstelter J, Bommireddy A, Choudhury A, Krup AL, Kaur P, et al. Smoothened-activating lipids drive resistance to CDK4/6 inhibition in Hedgehog-associated medulloblastoma cells and preclinical models. J Clin Invest. 2021;131(6).\u003c/li\u003e\n\u003cli\u003eGwynne WD, Suk Y, Custers S, Mikolajewicz N, Chan JK, Zador Z, et al. Cancer-selective metabolic vulnerabilities in MYC-amplified medulloblastoma. Cancer Cell. 2022;40(12):1488-502 e7.\u003c/li\u003e\n\u003cli\u003eWolpaw AJ, Bayliss R, Buchel G, Dang CV, Eilers M, Gustafson WC, et al. Drugging the \u0026quot;Undruggable\u0026quot; MYCN Oncogenic Transcription Factor: Overcoming Previous Obstacles to Impact Childhood Cancers. Cancer Res. 2021;81(7):1627-32.\u003c/li\u003e\n\u003cli\u003eTaylor L, Wade PK, Johnson JEC, Aldighieri M, Morlando S, Di Leva G, et al. Drug Resistance in Medulloblastoma Is Driven by YB-1, ABCB1 and a Seven-Gene Drug Signature. Cancers (Basel). 2023;15(4).\u003c/li\u003e\n\u003cli\u003eLi XY, Zhou LF, Gao LJ, Wei Y, Xu SF, Chen FY, et al. Cynanbungeigenin C and D, a pair of novel epimers from Cynanchum bungei, suppress hedgehog pathway-dependent medulloblastoma by blocking signaling at the level of Gli. Cancer Lett. 2018;420:195-207.\u003c/li\u003e\n\u003cli\u003ePhan TG, Croucher PI. The dormant cancer cell life cycle. Nat Rev Cancer. 2020;20(7):398-411.\u003c/li\u003e\n\u003cli\u003eLi C, Teixeira AF, Zhu HJ, Ten Dijke P. Cancer associated-fibroblast-derived exosomes in cancer progression. Mol Cancer. 2021;20(1):154.\u003c/li\u003e\n\u003cli\u003eSalsman VS, Chow KK, Shaffer DR, Kadikoy H, Li XN, Gerken C, et al. Crosstalk between medulloblastoma cells and endothelium triggers a strong chemotactic signal recruiting T lymphocytes to the tumor microenvironment. PLoS One. 2011;6(5):e20267.\u003c/li\u003e\n\u003cli\u003eLiang KH, Chang CC, Wu KS, Yu AL, Sung SY, Lee YY, et al. Notch signaling and natural killer cell infiltration in tumor tissues underlie medulloblastoma prognosis. Sci Rep. 2021;11(1):23282.\u003c/li\u003e\n\u003cli\u003eYildiz CB, Kundu T, Gehrmann J, Koesling J, Ravaei A, Wolff P, et al. EphrinA5 regulates cell motility by modulating Snhg15/DNA triplex-dependent targeting of DNMT1 to the Ncam1 promoter. Epigenetics Chromatin. 2023;16(1):42.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"medulloblastoma, chemotherapy resistance, single-cell RNA-seq, WNT, SHH, predictive model","lastPublishedDoi":"10.21203/rs.3.rs-7201592/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7201592/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eMedulloblastoma is a common primary tumor of the central nervous system. The impact of cellular heterogeneity on its treatment remains elusive.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eSingle-cell variational inference (scVI) model was used for batch effects correction. Gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA) were performed for evaluation of pathway activity. Cellchat algorithm was performed for inference of cell-cell interaction. SCENIC algorithm was performed for inferring gene regulatory networks (GRNs). Logistic regression and least absolute shrinkage and selection operator (LASSO) were conducted for identifying gene signature-associated with poor prognosis.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eThis study integrates single-cell RNA sequencing data from 7 medulloblastoma samples, which exhibited satisfactory batch effect correction (Silhouette batch: 0.91; cLISI: 0.97) and biological conservation (bioconservation score: 0.62) performance. Unsupervised leiden clustering identified 24 cellular clusters, including differentiated malignant cells, stem-like, proliferative, stress-responsive, immune cells, and cancer-associated fibroblasts. WNT (C6: \u003cem\u003eCTNNB1\u003c/em\u003e, \u003cem\u003eTSPYL1\u003c/em\u003e) and SHH (C14/C23: \u003cem\u003eATOH1\u003c/em\u003e, \u003cem\u003eSOX2\u003c/em\u003e) malignant cells exhibited pathway-specific enrichments. GSVA and GSEA implicated the activation of WNT and Hedgehog signaling pathways and overexpression of \u003cem\u003eMYCN\u003c/em\u003e, \u003cem\u003eABCB1\u003c/em\u003e, and \u003cem\u003eGLI\u003c/em\u003e in C14 SHH malignant cells. CellChat analysis revealed C14 SHH cells engage in ligand-receptor crosstalk (MIF, MDK, NCAM1) with immune/malignant cells, while SCENIC uncovered a regulatory network driven by SOX9, JUN/JUND, and SOX2, modulating inflammation, hypoxia, and WNT pathways. A LASSO-Cox integrated analysis identified a 13-gene signature (C14 signature) predicted poor prognosis (log-rank p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Functional enrichment analysis linked the signature to neurodevelopmental dysregulation and synaptic signaling.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThese findings demonstrate novel gene signature and cell subtype as potential driver of unfavorable prognosis, providing mechanistic insights and actionable biomarkers for medulloblastoma stratification.\u003c/p\u003e","manuscriptTitle":"Single-cell RNA-seq reveals multimodal regulatory networks and clinical predictive models in specific medulloblastoma cells","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-02 07:45:11","doi":"10.21203/rs.3.rs-7201592/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"dc584d7e-17c8-413b-a023-05a560b7605f","owner":[],"postedDate":"September 2nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-09-29T06:39:00+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-02 07:45:11","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7201592","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7201592","identity":"rs-7201592","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}