NCI Office of Data Sharing (ODS) Data Jamboree (Abstract Submissions): Submission #46

Presenter Information
---------------------






First Name: Jinghui









Middle Initial: {Empty}









Last Name: Zhang









Degree(s): PhD









Position/Title/Career Status: Professor









Organization: St Jude Children's Research Hospital









Organization Address:
Memphis










Email: jinghui.zhang@stjude.org









Other (Please Specify): {Empty}













Abstract Information
--------------------






Abstract Category: Solving a challenging clinical problem by data integration









Abstract Keywords: Ontology Classifier Data Integration









Abstract Title: Exploring new data integration approaches to enhance ontology-based pediatric cancer classification on challenging clinical cases 









Abstract Summary:
Ontologies designed for cancer classification have redefined our understanding of cancer by providing a hierarchical structure of complex biomedical data. Integration of omics findings with other established approaches such as histology and immunohistochemistry is now the new standard for clinical practice as reflected in the 5th edition of WHO CNS tumor classification scheme. We have developed a pediatric cancer focused ontology framework by leveraging existing efforts from OncoTree, WHO, and community knowledge. This framework has been applied to >5,000 pediatric samples with omics data accessible on the St Jude Cloud platform (https://www.stjude.cloud/) and more recently on >1,000 pediatric solid and CNS tumors profiled by the Childhood Cancer Data Initiative (CCDI) based on the clinical annotation as well as genomic alterations identified from exome and RNA-seq/Archer fusion platform.  

For CCDI sample classification, we have encountered multiple challenging cases with ambiguous or conflicting results indicating additional analytical approaches or data may improve classification. For example, for samples annotated as “Small round blue cell tumor” that have bi-allelic loss of SMARCB1 identified by exome analysis, can they be classified as rhabdoid tumors if additional information on tissue source or clinical imaging data can be obtained? Can newly developed RNA-seq expression-based machine-learning approaches, such as CanID (https://github.com/chenlab-sj/CanID), be used to augment classification over biomarker-based approaches? Can genome-wide copy number profile be used to improve the classification? 

Our project is aimed at exploring the value of new data and new analyses in advancing pediatric cancer classification. We will select ~20 challenging CCDI cases for real-time analysis at the Jamboree with a team of algorithm developers, ontology designers and clinical analysts to gain insights on future directions to improve the precision of ontology-based classification.










Upload Abstract: {Empty}