Childhood Cancer Data Initiative Annual Symposium (Abstract Registration): Submission #54

serial: '54'
sid: '150456'
uuid: e937c80f-e97b-46cf-a8ab-e1de3537311d
uri: /nci/ccdisymposium/abstract
created: '1756626984'
completed: '1756627232'
changed: '1756627232'
in_draft: '0'
current_page: ''
remote_addr: 10.208.24.230
uid: '0'
langcode: en
webform_id: ccdi_symposium_abstract
entity_type: node
entity_id: '2139'
locked: '0'
sticky: '0'
notes: ''
metatag: meta
data:
  authors_:
    - add_author_degree: M.S.
      add_author_first_name: Aishwarya
      add_author_last_name: Budhkar
      add_author_middle: ''
      add_author_organization: 'Indiana University Bloomington'
    - add_author_degree: M.S.
      add_author_first_name: Zanyu
      add_author_last_name: Shi
      add_author_middle: ''
      add_author_organization: 'Indiana University Indianapolis'
    - add_author_degree: Ph.D.
      add_author_first_name: Karen
      add_author_last_name: Pollok
      add_author_middle: E.
      add_author_organization: 'Indiana University School of Medicine, IU Simon Comprehensive Cancer Center'
    - add_author_degree: Ph.D.
      add_author_first_name: Kun
      add_author_last_name: Huang
      add_author_middle: ''
      add_author_organization: 'Indiana University School of Medicine, IU Simon Comprehensive Cancer Center'
    - add_author_degree: Ph.D.
      add_author_first_name: Jing
      add_author_last_name: Su
      add_author_middle: ''
      add_author_organization: 'Indiana University School of Medicine, IU Simon Comprehensive Cancer Center'
  abstract: |-
    The Childhood Cancer Data Initiative (CCDI) strives to accelerate pediatric cancer research by uniting clinical records with emerging spatial omics data but lacks a unified framework for joint analysis. Although CCDI and public repositories provide rich patient records and high-resolution spatial profiles, no unified data standards and framework exist to extract, harmonize, and process these multimodal datasets for interpretable discovery. 
    We developed the GAIPO (Graph Artificial Intelligence for Pediatric Oncology) clinicospatial data standard and platform, enabling graph artificial intelligence (AI) on the CCDI ecosystem. The GAIPO clinicospatial standard unifies spatial assay formats, spatial resolutions, and registration of spatial data across modalities. To promote interoperability and reproducibility, we define a spatial data model that encompasses data standards, formats, and interfaces to integrate imaging and omics data through spatial graph representation, thereby enabling the implementation of explainable graph AI models. The GAIPO platform provides: flexible interfaces to CCDI APIs and public portals for retrieval of clinical metadata, spatial omics data, and associated histopathology images; a preprocessing pipeline for omics and imaging data; a graph AI module that integrates multiple modalities for downstream analysis; and an interpretation module to identify important features and cells and reveal critical enrichment pathways and cell-cell interactions. We demonstrated GAIPO’s functionalities by using xSiGra as the explainable graph AI model, graph Grad-CAM as the interpretability module, ScPCA as the CCDI-participating resource, and Wilms tumor as the pediatric cancer. 
    The GAIPO standards and platform facilitate graph AI on clinicospatial data in precision oncology for pediatric cancers.
  abstract_title_: 'GAIPO introduces explainable graph AI for clinicospatial data in precision oncology of pediatric cancer'
  email_address_: abudhkar@iu.edu
  institution_: 'Indiana University Bloomington'
  presenting_author_: 'Aishwarya Budhkar'