Journal: Cell Genomics

Article Title: ChromBERT: A foundation model for learning interpretable representations for context-specific transcriptional regulatory networks

doi: 10.1016/j.xgen.2025.101130

Figure Lengend Snippet:

Article Snippet: PBMC 10× single-cell multi-omics dataset , 10× Genomics Datasets , https://www.10xgenomics.com/datasets/pbmc-from-a-healthy-donor-granulocytes-removed-through-cell-sorting-10-k-1-standard-2-0-0.

Techniques: Gene Expression, Single Cell, Biomarker Discovery, Software

Journal: iScience

Article Title: Resource: A compendium of HLA types and expression in pediatric cancer models

doi: 10.1016/j.isci.2025.113887

Figure Lengend Snippet: Computational workflow and data integration for HLA expression, typing, and neoepitope prediction in the CCMA (A) WGS and RNA-Seq data from 287 cell lines were processed for GSVA and Class I HLA typing. HLA pathway activity scores were compared to primary tumors (OpenPedCan). The cell line-specific somatic mutations and HLA types were processed to predict potentially effective neoepitopes (Figure created with BioRender.com ). (B) Decision-making process for selecting the optimal HLA types. The final HLA types were determined hierarchically by each criterion based on data availability. HLA types inferred from Optitype were denoted by the dash-line box. See also .

Article Snippet: CCMA primary multi-omics datasets , Mendeley Data , https://doi.org/10.17632/rnfs539pfw.1.

Techniques: Expressing, RNA Sequencing, Immunopeptidomics, Activity Assay

Journal: iScience

Article Title: Resource: A compendium of HLA types and expression in pediatric cancer models

doi: 10.1016/j.isci.2025.113887

Figure Lengend Snippet: Tile plot for the CCMA HLA types Tile plot shows data integration, supertypes, and zygosity for the CCMA models. Subplots were grouped by cancer type and clustered by CCMA cancer class. Data, including result type (high-resolution typing, match, partial match, manual curation, and single result), supertypes (“One” representing allele and “Two” representing alleles), and zygosity (homozygous) for each model, were color-coded as indicated. See also and .

Article Snippet: CCMA primary multi-omics datasets , Mendeley Data , https://doi.org/10.17632/rnfs539pfw.1.

Techniques:

Journal: iScience

Article Title: Resource: A compendium of HLA types and expression in pediatric cancer models

doi: 10.1016/j.isci.2025.113887

Figure Lengend Snippet: Distribution of HLA supertypes, zygosity, and ASE loss events across the CCMA (A) Stacked bar chart shows the HLA allele count of CCMA models across 12 supertypes or unclassified. HLA supertypes were defined according to Sidney et al., 2008. (B) Comparison of HLA class I homozygosity percentage and ASE across three genes in cancer types. HLA class I ASE is represented by the minimal raw minor allele frequency (MAF) among the three HLA class I genes. A reference line MAF = 0.25 was labeled in red. See also .

Article Snippet: CCMA primary multi-omics datasets , Mendeley Data , https://doi.org/10.17632/rnfs539pfw.1.

Techniques: Comparison, Labeling

Journal: iScience

Article Title: Resource: A compendium of HLA types and expression in pediatric cancer models

doi: 10.1016/j.isci.2025.113887

Figure Lengend Snippet: Predicted neoepitope profiles for CCMA (A) Subgroup intersections between filtering criteria for predicted neoepitopes. Thresholds were determined by three types of prediction algorithms, favoring different epitope qualities (binding and elution). For binding and elution criteria, weak binders (IC50 < 500 nM; percentile ranking <2%), and strong binders (IC50 < 50 nM; percentile ranking <0.5%) were considered. (B) Distribution of predicted neoepitopes with IC50 < 500 nM across 11 CCMA cancer classes. The median number of neoepitopes for each class was labeled. (C) Top 20 recurrently mutated genes within the predicted neoepitope catalog. Distribution of the genes across different cancer types was shown as labeled. See also and .

Article Snippet: CCMA primary multi-omics datasets , Mendeley Data , https://doi.org/10.17632/rnfs539pfw.1.

Techniques: Binding Assay, Labeling

Journal: iScience

Article Title: Resource: A compendium of HLA types and expression in pediatric cancer models

doi: 10.1016/j.isci.2025.113887

Figure Lengend Snippet: Transcriptional profiling of HLA and antigen presentation genes in the CCMA Cohort (A) Schematic plot shows transcriptional profiles of cellular pathways and key regulators in HLA and antigen presentation. (B) Violin plots show the expression of individual HLA genes ( HLA-A , HLA-B , and HLA-C ) across CCMA cancer types. No significant differences were identified using one-way ANOVA. (C) Aggregated pathway boxplot categorized by cancer types, highlighting differential immune-related pathway activity. (D) Correlation plot illustrates significant relationships among APP, HLA class I, HLA class II, and proteasome pathways. (E and F) Heatmaps showing pathway activity scores for key immune-related and proteasome pathways across distinct subtypes of H3K27M-DMG (E) and ATRT (F), namely DMG-H3.1K27M and DMG-H3.3K27M, Group 1 and Group 2 ATRT. Each row represents a specific pathway, and each column corresponds to an individual tumor sample. Color scheme indicates relative expression levels, red indicates increased activity, white neutral expression, and blue downregulation. Hierarchical clustering, using Ward’s method with Euclidean distance. ∗ Denotes FDR-adjusted p < 0.05 assessed using a two-tailed Student’s t test or Wilcoxon rank-sum test, with multiple testing correction via the Benjamini-Hochberg procedure. See also and , and .

Article Snippet: CCMA primary multi-omics datasets , Mendeley Data , https://doi.org/10.17632/rnfs539pfw.1.

Techniques: Immunopeptidomics, Expressing, Activity Assay, Two Tailed Test

Journal: iScience

Article Title: Resource: A compendium of HLA types and expression in pediatric cancer models

doi: 10.1016/j.isci.2025.113887

Figure Lengend Snippet: HLA class I gene expression across pediatric cancer subtypes in different sample types (A–C) Boxplots display log2-transformed transcript per million (TPM) expression values of HLA-A (A), HLA-B (B), and HLA-C (C) across various pediatric cancer subtypes in solid tumor samples (green) and derived cell lines (purple) from OpenPedCan, and CCMA cell lines (orange) from our analysis. Statistical significance was determined by a two-way ANOVA test ( p < 0.001). Data are presented as boxplots showing the median (center line), interquartile range (box), and whiskers extending to 1.5 × the interquartile range; individual points represent outliers. See also and and .

Article Snippet: CCMA primary multi-omics datasets , Mendeley Data , https://doi.org/10.17632/rnfs539pfw.1.

Techniques: Gene Expression, Transformation Assay, Expressing, Derivative Assay

Journal: BMC bioinformatics

Article Title: Combining single-cell ATAC and RNA sequencing for supervised cell annotation.

doi: 10.1186/s12859-025-06084-6

Figure Lengend Snippet: Fig. 1 Overview of supervised annotation workflow. Fig. 1 illustrates an overview of the supervised annotation workflow, which involved three primary steps. Initially, a 10x Genomics multi-omic (RNA + ATAC) sequencing dataset detailing peripheral blood mononuclear cells (PBMC) or neuronal cells underwent rigorous quality control (QC) procedures. Subsequently, ground truth labels were derived through a pre-processing pipeline, followed by weighted nearest neighbors (WNN) clustering and marker gene-based annotation. These labels served as the ground truth for the supervised classification models. The final step involved generating 10 bootstrapped train and out-of-bag (OOB) test sets. This was followed by pre-processing and dimensionality reduction using principal component analysis (PCA) or single-cell Variational Inference (scVI). Classification was then performed using support vector machine (SVM), random forest (RF), or logistic regression (LR) models. Models utilizing RNA-only embeddings were compared to those utilizing RNA and ATAC

Article Snippet: Data availability The 10x multi-omics PBMC dataset used in this study is available on the 10x Genomics website: https:// www.10xgenomics.com/resources/datasets.

Techniques: Sequencing, Control, Derivative Assay, Marker, Plasmid Preparation

Journal: BMC Bioinformatics

Article Title: Combining single-cell ATAC and RNA sequencing for supervised cell annotation

doi: 10.1186/s12859-025-06084-6

Figure Lengend Snippet: Overview of supervised annotation workflow. Fig. 1 illustrates an overview of the supervised annotation workflow, which involved three primary steps. Initially, a 10x Genomics multi-omic (RNA + ATAC) sequencing dataset detailing peripheral blood mononuclear cells (PBMC) or neuronal cells underwent rigorous quality control (QC) procedures. Subsequently, ground truth labels were derived through a pre-processing pipeline, followed by weighted nearest neighbors (WNN) clustering and marker gene-based annotation. These labels served as the ground truth for the supervised classification models. The final step involved generating 10 bootstrapped train and out-of-bag (OOB) test sets. This was followed by pre-processing and dimensionality reduction using principal component analysis (PCA) or single-cell Variational Inference (scVI). Classification was then performed using support vector machine (SVM), random forest (RF), or logistic regression (LR) models. Models utilizing RNA-only embeddings were compared to those utilizing RNA and ATAC

Article Snippet: The 10x multi-omics PBMC dataset used in this study is available on the 10x Genomics website: https://www.10xgenomics.com/resources/datasets.

Techniques: Sequencing, Control, Derivative Assay, Marker, Plasmid Preparation