Poster #61 - Deanne Taylor

Ontology‑Aware Integration of Kids First and dGTEx data for Pediatric Anatomy

Deanne Taylor PhD (CHOP/UPenn), Adam C Resnick PhD (CHOP/UPenn), Thomas Bell PhD (NDRI), Kristin Ardlie PhD (Broad Institute), Rebecca Linn MD (CHOP/UPenn), Raquel Hernandez MD, MPH (JHU), Melissa VonDran PhD (NDRI) Emilie Hattrell (NDRI), Patrick Van Hoose (NDRI), Thomas Blanchard PhD (U. Maryland), Brady Walker (Broad Institute), Jared Nedzel PhD (Broad Institute), Nenad Sestan MD PhD (Yale), Jonathan Silverstein MD (UPitt) Alan Simmons (UPitt), Katy Borner PhD (Indiana University), Andreas Bueckle PhD (Indiana University), Sharon Diskin PhD (CHOP.UPenn), Jo Lynne Rokita (Children's National, George Washington University) Taha Mohseni Ahooyi PhD (CHOP), Ryan James Corbett PhD (Children's National), Aditya Lahiri (CHOP), Asif Chinwalla PhD (CHOP), Dave D. Hill (CHOP), Shiping Zhang PhD (CHOP), Benjamin J Stear (CHOP)

The Developmental Genotype Tissue Expression (dGTEx) program is creating a reference atlas of gene expression across healthy pediatric tissues. Its sample data dictionary defines 51 tissue‑site details grouped into 24 Uberon organ systems. This framework will allow transcriptomic data to be anchored to anatomy and compared across childhood ages. The Gabriella Miller Kids First Pediatric Research Program (Kids First) builds genomic and clinical resources for childhood cancer and structural birth defects, with the Kids First Data Resource Center (KFDRC) providing access to germline variation and phenotypes. The Data Distillery Knowledge Graph (DDKG) links these resources with more than 180 biomedical ontologies, including Uberon, MONDO, and HPO. Today, the DDKG provides a platform where Kids First and dGTEx data can be analyzed in an ontology‑aware environment. Extension of the HuBMAP Human Reference Atlas (HRA) into a developmental HRA (dHRA) will provide the anatomical framework needed to compute across these resources in a shared spatial context. We evaluated overlap between 667 unique Kids First phenotypes from birth defect cohorts and a 34‑donor dGTEx pilot dataset that is currently available within the dGTEx consortium and planned for public release in the coming months. This pilot includes 342 RNA‑seq samples across 51 tissue sites and 24 organs. Using the dGTEx‑to‑Uberon key, each sample was assigned to an organ. We curated 667 unique Kids First phenotypes (HPO terms), linked them to Uberon tissues, and measured coverage as the share of phenotypes that matched at least one profiled organ or tissue site with at least one pilot sample. We mapped dGTEx tissue site details to organ level using the dGTEx Uberon key, and linked Kids First HPO phenotypes to Uberon tissues with a curated table, deduplicated by HPO ID. Coverage was defined as the proportion of phenotypes whose mapped Uberon term matched at least one organ represented in the pilot. We found that 27.7% of Kids First phenotypes mapped to at least one profiled dGTEx organ/tissue in the pilot data (185/667). The most common overlaps involved cardiovascular, renal, pulmonary, digestive, and integumentary systems, while clear gaps remained for eye, ear, bladder/urinary, craniofacial, bone/skeletal, and marrow/hematologic phenotypes, with the missing majority being skeletal. The current dGTEx pilot already provides useful context for many Kids First phenotypes, even before public release. Embedding dGTEx and Kids First within the DDKG ensures these resources are interoperable today. Extending the HRA into the dHRA will supply the anatomical reference needed to integrate these datasets at scale. Together, they will form a pediatric data integration system that connects healthy tissue expression, germline variation, and phenotype data within a computable anatomical framework.