GDC demo shows Correlation Plot tool for linking mutations, CNV, expression and clinical outcomes

Bill, director of user services at the University of Chicago, opened a Genomic Data Commons webinar and said the session would demonstrate the GDC’s new Correlation Plot tool for exploring relationships among clinical metadata, somatic mutations, copy-number calls and gene expression.

"The webinar will be recorded," Bill said, and noted slides and a transcript will be posted on the GDC website. He described two main data classes the tool uses: directly submitted case and clinical metadata (for example, primary site, disease type, demographics and family history) and GDC-derived molecular measures such as masked somatic mutation calls, gene-level copy-number states and FPKM-UQ expression values produced by the GDC pipelines.

Xin, Bill’s colleague, then launched a live demo from the GDC portal’s Analysis Center and walked through quick-access comparisons and the general Correlation input panel. Using an IDH1 mutation vs. disease-type example restricted to brain and breast primary sites, he showed the IDH1 mutation concentrated in glioma cases and said the plot tooltip reported about "38.1%" (with a tooltip case count shown in the interface) for mutated cases in that view. He then used the plot’s "list cases" feature to display case submitter IDs and the exact mutation annotations (for example, R132H and other missense variants).

Xin demonstrated a multi-gene (gene-set) mutation query by submitting IDH1, ATRX and PTEN. He noted the current tool limits gene-set queries to 10 genes for performance reasons; in his example the aggregated gene-set increased the plotted mutation proportion (near 60% among glioma cases in his view).

For copy-number examples, Xin used EGFR and showed amplification calls concentrated in gliomas. He then produced Kaplan–Meier survival plots comparing mutated versus wild-type cases (IDH1) and showed that stratifying by primary site produced the strongest survival separation in brain cases. In a PTEN example, Xin pointed out samples with a heterozygous deletion plus a PTEN mutation — a putative "double hit" pattern concentrated in brain samples — and demonstrated listing those cases for per-sample mutation and CNV details.

Xin also demonstrated expression analyses: EGFR expression stratified by CNV category (amplified samples had higher EGFR expression) and expression-based survival comparison using a default cutoff (6.82) and a custom cutoff workflow (he entered 10 and 100 to make three bins). He described the observed trend in the demo as higher EGFR expression associated with poorer outcome in the plotted cohort.

Bill closed the demonstration by showing where to find the Correlation Plot card (Analysis Center) and the user documentation (docs.gdc.cancer.gov). He reminded attendees that the portal hosts harmonized TCGA data (the cohort selector example listed ~11,428 TCGA cases in the portal view) and that controlled-access files require dbGaP authorization; the demo used only open-access files. Bill pointed users to pipeline documentation (DNA-seq, variant calling, RNA-seq) and release notes for details on harmonization and new data releases.

The presenters answered live questions about how GDC relates to TCGA and cBioPortal and confirmed GDC’s harmonization enables cross-study comparison via standard pipelines. They closed by promising to post the recording and slides and provided a support contact (support@nci-gdc.datacommons.io).