Data Sharing and Integrative Analysis Core

点击翻译按钮获取中文摘要

This Data Sharing and Integrative Analysis (DSIA) Core will support the overall mission of the Center through three interlocking functions: (Aim 1) to ensure effective data quality and full computability, (Aim 2) to provide innovative integrative analyses to support the scientific goals of this ROBIN center, and (Aim 3) to ensure seamless data sharing for inter-ROBIN network collaborations as well as to the NCI Cancer Research Data Ecosystem. Aim 1. To ensure effective data quality and computability. Under this Aim, we will collect, harmonize, and make accessible the digital data collected in this ROBIN Center, including the Molecular Characterization Trial, as well as Projects 1 and 2. These diverse sets of data include: 1) clinical data (de-identified patient and tumor characteristics, Immunoscore of the diagnostic rectal cancer specimen, etc.), imaging data (MRI / CT images at baseline and prior to surgery), 2) radiotherapy treatment planning data (DICOM images, dose distributions to the contoured anatomic structures), and 3) biological data resulting from the two scientific projects associated with the MCT (e.g., spatial transcriptomics, microbiome, immune biomarkers in circulating blood, etc.). We will curate and transfer data from the Molecular Characterization Trial (MCT) and Scientific Projects 1 and 2, and will fully link all ROBIN data within NCI Cloud Resource FireCloud workspaces and the Imaging Data Commons, with imputation where necessary, providing fully computable subject data profiles. Aim 2. To conduct innovative, integrative analyses to support the scientific goals of this ROBIN center. Under this Aim, we will apply both unbiased/non-parametric and machine learning integrative (multi-datatype) analyses to identify critical immune phenotypes and their tumor/immune molecular signatures using the full spectrum of available biological and imaging data. To identify biological and imaging/radiomics signatures or subtypes, we will apply innovative clustering using network optimal mass transport methods. To understand the impact of radiation on Peripheral Blood Mononuclear Cells (PBMCs), we will conduct systematic multivariate analyses using machine learning approaches. To understand subtypes of RT response, we will apply a novel non-linear machine-learning integrative phenotypic mapping tool (iPhenMap), based on sparse Bayesian factor analysis modeling, that integrates molecular and functional multimodal patterns. Aim 3. To support seamless data sharing for inter- ROBIN network collaborations and cross-training. Under this Aim, we will document and demonstrate our tools, data, and rerunnable analysis workflows in FireCloud and Imaging Data Commons infrastructure, to support inter-ROBIN network collaborations, as well as inter-disciplinary cross-training.