Cohorts and projects
A project defines data provenance, design, tumour context and available cases. Different projects cannot be merged mechanically.
NCI GDC / RESEARCH LAYER 0.14
We show which tumour and molecular data are available in NCI GDC, how to read aggregate frequency and why it cannot become a personal prognosis or treatment recommendation.
SIX LAYERS
Each layer has its own specimen, method, observation unit and limitations.
A project defines data provenance, design, tumour context and available cases. Different projects cannot be merged mechanically.
Open MAF data support counts of SSM observations in genes and variants. A gene record does not establish driver status, assay fitness or a drug target.
These are separate data classes and analysis methods. SSM frequency does not describe amplification, deletion, fusion or rearrangement.
RNA-seq and methylation answer different research questions. High expression is not an activating variant and does not create a treatment indication.
Completeness, definitions, missingness and treatment era vary. Observed cohort survival is not an individual prognosis.
Open data require no authorisation; controlled access follows participant consent and separate authorisation. The atlas does not request or bypass controlled access.
INTEGRATION BOUNDARY
Which cases are actually in the denominator and have the required data type?
Is the unit a gene, exact variant, consequence class, CNV, expression or another entity?
How were specimens selected and which technologies and pipelines were used?
Is there clinical confirmation, a regulatory indication and a diagnostically validated assay?
Are we mistaking a research association or cohort survival for causality and individual prognosis?