Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.31816

TAPESTRY: an interactive resource for exploring tumor-enriched splice junctions in pediatric brain tumors

Patricia J. Sullivan·Ryan J. Corbett·Chao Di·Alex Sickler·Aylar Babaei·Meghali Goswami·Jo Lynne Rokita
View original source ↗Read source document ↗Export metadata ↓

Abstract

TAPESTRY (https://tapestry.rokitalab.com) is a database and web application for exploring recurrent tumor-enriched splice junctions (TEJs) in pediatric brain tumors, cancers that carry few protein-coding mutations and therefore few mutation-derived neoantigens. However, TEJs represent a much larger source of neoantigens as potential therapeutic targets in these tumors. TAPESTRY catalogs 45,430 recurrent TEJs across 19 histologies, identified in 1,962 RNA-seq specimens (including 1,272 independent primary tumors) from the Pediatric Brain Tumor Atlas and compared to four normal brain reference cohorts comprising adult, pediatric, developmental and cell-type-resolved tissue. Each junction is classified as tumor-specific or oncofetal; annotated for splice-event type, coding consequence, novelty relative to GENCODE v39 and overlap with Pfam and UniProt protein domains; and scored for recurrence within each histology. Users can search by gene or splicing event and filter results by histology, junction class, and annotation. Users can also generate gene-level expression heatmaps across cohorts, junction-level expression boxplots, and transcript and exon diagrams, and download all tables and figures. A documented REST API supports programmatic access, and all source code is openly available. By making tumor-selective and developmentally reactivated splicing events searchable and interpretable, TAPESTRY supports the prioritization of splice-derived targets for antibody-based, cell-based and vaccine immunotherapies in pediatric brain and central nervous system tumors.

Explore related subjects

q-bio.GN

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Patricia J. Sullivan, Ryan J. Corbett, Chao Di, Alex Sickler, Aylar Babaei, Meghali Goswami, Jo Lynne Rokita. 2026-10-22. TAPESTRY: an interactive resource for exploring tumor-enriched splice junctions in pediatric brain tumors. https://arxiv.org/abs/2610.31816

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

OmicsLM: A Multimodal Large Language Model for Multi-Sample Omics Reasoning

Large language models can reason over biological knowledge, but they are not naturally equipped to work with quantitative omics measurements. Existing approaches either convert expression profiles into long textual representations, often losing quantitative information, or connect omics encoders to language models for narrow tasks. We introduce OmicsLM, a 9B multimodal language model that represents each bulk or single-cell transcriptome as a compact continuous embedding that can be interleaved with natural-language instructions and with other samples in the same context. An key contribution is a training corpus that connects experimental measurements with biological knowledge and task supervision. We combine templates linking transcriptomic profiles to annotations and experimental outcomes, reasoning examples connecting evidence across sources, and LLM-generated supervision with task-specific quality checks. The instruction-tuning corpus spans 157 biological and general-purpose task groups. Training includes continued pretraining on 20 billion tokens and supervised fine-tuning on 5.21 million conversations. We further introduce GEO-OmicsQA, a publication-disjoint benchmark of 3,000 questions requiring models to interpret and compare expression profiles from real GEO studies. OmicsLM achieves the best results among the evaluated models on three of four held-out annotation tasks and on all six GEO-OmicsQA variants, and the best average performance among compared methods on PerturbQA perturbation-response prediction. Ablations show a clear advantage for continuous omics representations over a text-only gene-list representation. Together, these results show that quantitative transcriptomic measurements and biological language can be integrated within a general conversational model, enabling both predictive tasks and language-guided interpretation of experimental samples.

q-bio.GN↗
arXiv2026

Benchmarking miRNA Target Prediction Methods

Background: MicroRNAs (miRNAs) regulate gene expression by binding to target mRNAs. Although many tools predict miRNA-target interactions, predictions show limited concordance, and benchmarks relying on curated databases may be biased by overlap with tool-development data. Methods: We benchmarked seven tools (TargetScan, miRDB, miRmap, DIANA-microT, RNA22, miRWalk, and mirDIP) across approximately 59.7 million candidate miRNA-mRNA pairs. Each pair received an evidence-graded reference score (0-3) combining validated interactions from miRTarBase and TarBase with expression-derived evidence from paired miRNA-seq and mRNA-seq data in an induced pluripotent stem cell (iPSC) cohort of 22q11.2 deletion syndrome and control lines. Performance was evaluated using Type I error, statistical power, and false discovery proportion (FDP). Robustness was assessed by increasing reference-score stringency (truth signal) and including progressively lower-confidence predictions (model signal). Optimal cutoffs maximized Youden's J statistic. Results: Rankings were largely consistent across settings. TargetScan and miRDB maintained Type I error rates below 0.05, with TargetScan showing the lowest FDP, but both had limited power. mirDIP achieved the highest power at the cost of the highest Type I error rate. At optimal cutoffs, mirDIP best balanced sensitivity and specificity (Youden's J = 0.566), followed by DIANA-microT (J = 0.481); miRWalk and RNA22 had FDPs near the random-prediction baseline. No tool was optimal across all metrics; tool and cutoff selection should reflect whether sensitivity or specificity is prioritized."""

q-bio.GN↗
arXiv2026

Benchmarking Recent Single-Cell Synthetic Data Generation Methods on the OneK1K Dataset

Synthetic single-cell RNA-seq data generation offers a promising route for privacy-aware biomedical data sharing, but donor-linked datasets require careful balancing of biological utility and privacy risk. In this work, we benchmark synthetic generation methods for the CAMDA 2026 ELSA Health Privacy Challenge using the donor-split OneK1K dataset, comparing scGFT, scLDM, and Cell2Sentence across statistical fidelity, downstream biological preservation, and membership inference vulnerability. Because full-transcriptome generation is computationally demanding, we focus generative modeling on highly variable genes and evaluate alternative strategies for reconstructing non-HVG expression. Our results show that non-HVG handling is a central determinant of privacy: directly copying non-HVG values leads to severe leakage, whereas Poisson-based reconstruction substantially reduces membership inference risk. We further find that scGFT and scLDM better preserve biological structure than an independent Poisson baseline, while maintaining near-random privacy profiles under appropriate configurations. These findings highlight the importance of evaluating synthetic single-cell data beyond marginal similarity alone, emphasizing robustness, interpretability, and privacy-conscious design choices throughout preprocessing, generation, and post-processing.

q-bio.GN↗