Research
Below are some of our present and past research projects.
Tool and Resource Development
We develop open source resources that can drive innovation and accelerate the discovery of novel therapeutic strategies for pediatric brain tumor research.
TAPESTRY: Tumor Alternative PEdiatric Splicing visualizaTion and queRYing (TAPESTRY) is our latest web application and API for exploring tumor-enriched junctions (TEJs) across pediatric brain tumors. TAPESTRY was developed by Dr. Patricia Sullivan, with the underlying tumor-enriched junction data generated by Dr. Ryan Corbett, and AWS infrastructure support from Dr. Chao Di. Explore TAPESTRY here.
OpenPedCanExpress: OpenPedCanExpress is a lightweight web application for exploring gene expression in pediatric tumors versus normal tissue controls, built on OpenPedCan data and using TAPESTRY plot groups and colors. Search any gene, filter by cohort, histology, and molecular subtype, and export publication-ready plots.
OpenPedCan: The pan-cancer expansion counterpart of OpenPBTA is the Open Pediatric Cancer (OpenPedCan) Project. OpenPedCan became the foundational data layer of the NCI’s pediatric Molecular Targets Platform and is used to support clinical trial decision-making for the Pediatric Neuro-Oncology Consortium (PNOC), therapeutic development, and discovery and translational research.
OpenPBTA: With Alex’s Lemonade Stand Foundation’s Childhood Cancer Data Lab, Dr. Rokita co-led the creation of the Open Pediatric Brain Tumor Atlas (OpenPBTA), an open-source reproducible analysis platform consisting of nearly 50 scalable modules to analyze pediatric brain tumor data.
AutoGVP: In collaboration with Dr. Sharon Diskin at CHOP, we created AutoGVP, a dockerized tool to assess germline variant pathogenicity based on current ACMG guidelines. We also created a pathogenicity preprocessing CAVATICA application to add ClinVar, InterVar, ANNOVAR, and AutoPVS1 annotations to the germline VCFs prior to running AutoGVP.
annoFuse: We created an R package, annoFuse, and companion RShiny application to prioritize putative oncogenic fusions and visually explore fusion calls and breakpoints across a cohort.
Medulloblastoma classifier: We developed a transcriptome-based classifier that assigns medulloblastoma samples to the four molecular subtypes (WNT, SHH, Group 3, and Group 4) using gene expression ratios, making it robust across RNA-seq and microarray platforms. It achieved a median accuracy of 97.8% across 15 independent studies comprising 1,286 samples. The classifier is available as an R package, medulloPackage, and the development code is in the medullo-classifier-dev repository.
Integrative Multi-omics Approach
Our research harnesses the power of integrative multi-omics to unravel the complexities of pediatric brain tumors. By combining data from genomics, transcriptomics, and proteomics, we uncover critical insights into the biology of these aggressive cancers.
RNA splicing and identification of novel targets: We investigate the role of RNA splicing in driving tumor formation and progression in pediatric brain tumors. We have developed a workflow to identify and prioritize splice-derived peptides as new candidate tumor-specific vulnerabilities, paving the way for personalized immunotherapeutic treatment strategies for pediatric brain tumor patients. Our work has identified CLK1 as a candidate oncogenic dependency in pediatric CNS tumors and an NRCAM variant defined by microexon skipping as a targetable cell surface proteoform in high-grade gliomas.
Germline risk variants: In collaboration with Dr. Sharon Diskin, our team characterizes patient and parental germlines to pinpoint pathogenic variants that predispose patients to brain tumor development, offering new insights into how these variants influence tumor biology.
Clinical Trial Support
We bring real-time bioinformatics to the clinic, analyzing patient tumor data to help guide personalized treatment strategies for children with brain tumors.
IMPACT: Immunotherapy for Malignant Pediatric Brain Tumors Employing Adoptive Cellular Therapy (IMPACT) is a phase 1 interventional study led by Dr. Brian R. Rood. It tests personalized tumor-specific antigen T cells (TSA-T), which target proteogenomically determined antigens from each patient’s own tumor, in children under 5 years of age with embryonal brain tumors. As a sub-investigator, Dr. Rokita led a team of bioinformatics scientists and engineers in developing and implementing the IMPACT bioinformatics workflows, which are available on GitHub.
SUPERNOVA-RBT: SUPERNOVA-RBT is a pilot interventional study led by Dr. Adriana Fonseca that tests real-time drug screening and genomic sequencing of tumor tissue taken during surgery, in children and young adults with relapsed ultra-rare brain tumors. A specialized tumor board uses the results to assign a personalized treatment plan. As a sub-investigator, Dr. Rokita oversaw the bioinformatic integration of genomic, transcriptomic, and drug-screen data, which places each tumor’s response to therapy in its molecular context.
Health Disparities
We are deeply passionate about exploring the impact of social determinants of health, race, ethnicity, and genetic-based ancestry on pediatric brain tumors. In collaboration with Dr. Cassie Kline, we recently identified significant associations between genetic ancestry superpopulations, prevalence of CNS tumor histologies and molecular subtypes, patient overall and event-free survival, and upfront treatment approaches.
Featured
A dockerized workflow integrating ClinVar and InterVar germline sequence variant classification
A harmonized open-source multi-omic dataset from > 6,000 pediatric cancer patients with > 7,000 tumor events across more than 100 histologies.
A global open science initiative characterizing over 1,000 pediatric brain tumors.
Distinct associations of genetic ancestry-based superpopulations exist within pediatric CNS tumor histologic and molecular subtypes and correlate with survival outcomes and treatment.
Characterization of aberrant splicing in pediatric central nervous system tumors reveals CLK1 as a candidate oncogenic dependency
More
A dockerized workflow integrating ClinVar and InterVar germline sequence variant classification
A harmonized open-source multi-omic dataset from > 6,000 pediatric cancer patients with > 7,000 tumor events across more than 100 histologies.
A global open science initiative characterizing over 1,000 pediatric brain tumors.
Distinct associations of genetic ancestry-based superpopulations exist within pediatric CNS tumor histologic and molecular subtypes and correlate with survival outcomes and treatment.
Characterization of aberrant splicing in pediatric central nervous system tumors reveals CLK1 as a candidate oncogenic dependency