TCGA Pan-Cancer Atlas: 10,000 tumours across 33 cancer types, classified by molecular features
The capstone of The Cancer Genome Atlas integrated DNA, RNA, protein and methylation data on about 10,000 tumours, showing that cell of origin dominates molecular classification but that some cancers regroup across organs.
The Pan-Cancer Atlas was a set of 27 papers in Cell Press journals in April 2018 summarising a decade of TCGA. The flagship classification paper (Hoadley et al.) integrated five data types on 9,759 tumours from 33 cancer types and identified 28 molecular clusters. Most clusters were dominated by tissue of origin, but squamous cancers from different organs grouped together, as did gastrointestinal adenocarcinomas and kidney cancers of different histologies.
Companion papers catalogued 299 driver genes and over 3,400 driver mutations (Bailey et al.), showed that 89% of tumours had at least one driver alteration in ten canonical signalling pathways and 57% had at least one potentially targetable alteration (Sanchez-Vega et al.), and characterised immune subtypes, oncogenic processes and cell-of-origin patterns.
TCGA data, freely available through the Genomic Data Commons, became the reference against which nearly every cancer genomics study is compared.
- 9,759 tumours from 33 cancer types integrated across mRNA, miRNA, DNA methylation, copy number and protein data
- 28 iCluster molecular subtypes; about two-thirds dominated by tissue of origin, with cross-tissue clusters for squamous and pan-gastrointestinal cancers
- Companion paper: 299 driver genes, of which about half were not previously in curated driver lists (Bailey et al.)
- Companion paper: 89% of tumours had at least one alteration in ten signalling pathways; 57% had a potentially actionable alteration (Sanchez-Vega et al.)
Cancers are defined as much by the tissue they come from as by the mutations they carry, which is why the same drug can work in one organ and fail in another with the same mutation. TCGA is the shared public dataset behind most modern biomarkers and target discovery.
- Primary, untreated tumours only; metastatic and post-treatment biology are under-represented
- Predominantly white US patients; ancestry diversity is limited
- Bulk tissue profiling averages over heterogeneity later revealed by single-cell and spatial methods
- Actionability estimates count alterations with any drug evidence, not proven clinical benefit
Pages like this
not linked directly; found by shared links- IdeaA standard evolvability score for every tumour
Shares Mutational signature, Tumour mutational burden (TMB), Tumour heterogeneity and clonal evolution, Whole-exome & whole-genome sequencing.
- IdeaPick the laboratory model that matches the patient, not the one to hand
Shares Broad Institute of MIT and Harvard, Trials do not represent the people who get cancer, Data silos, National Cancer Institute (NIH).
- IdeaPool every multi-sample tumour genome into one open evolution atlas
Shares Broad Institute of MIT and Harvard, Tumour heterogeneity and clonal evolution, Whole-exome & whole-genome sequencing, Data silos.
- InstitutionOntario Institute for Cancer Research
Shares Tumour heterogeneity and clonal evolution, Whole-exome & whole-genome sequencing, Data silos, Comprehensive genomic profiling.
- IdeaBank three spatially separate tumour blocks from every resection
Shares Tumour heterogeneity and clonal evolution, Whole-exome & whole-genome sequencing, Data silos, Comprehensive genomic profiling.
- InstitutionSkåne University Hospital / Lund University Cancer Centre
Shares PAM50 / intrinsic subtypes, RNA sequencing & expression profiling, Whole-exome & whole-genome sequencing.
- IdeaA global rapid tissue donation network for metastatic disease
Shares TCGA / NCI Genomic Data Commons, Whole-exome & whole-genome sequencing, Data silos.
- CompanyBostonGene
Shares RNA sequencing & expression profiling, Data silos, Comprehensive genomic profiling.