Integrated LUAD Data and Code

Integrated lung adenocarcinoma (LUAD) dataset and R pipeline for: (1) cancer effect size estimation for somatic variants, (2) variant × tumor-suppressor-gene epistasis testing, and (3) comparison of human variant selection strength with tumor burden in genetically engineered mouse models (GEMMs).

1. Data

1.1 Cohort origin and sequencing type

The integrated dataset merges LUAD cohorts spanning whole-genome (WGS), whole-exome (WES), and targeted/panel sequencing, downloaded from cBioPortal datahub or the NCI Genomic Data Commons, plus the Yale cohort from institutional records.

Folder	Source code	Sequencing	Study	Institution/project	n
data/genie_9/	Genie	Targeted (multiple panels)	AACR Project GENIE Consortium, 2017 [G1]	AACR Project GENIE v9.0.0	5,997
data/luad_fm-ad/	FM-AD	Targeted (FoundationOne CDx)	Hartmaier et al., 2017 [G2]	Foundation Medicine Adult Cancer Dataset (GDC FM-AD)	1,206
data/lung_msk_2017/	MSK2017	Targeted (IMPACT 341, 410)	Jordan et al., 2017 [G3]	MSKCC	372
data/nsclc_pd1_msk_2018/	MSK2018	Targeted (IMPACT 341, 410, 468)	Rizvi et al., 2018 [G4]	MSKCC	186
data/luad_tcga/	TCGA	WES	TCGA Research Network, 2014 [G5]	The Cancer Genome Atlas	515
data/luad_oncosg_2020/	OncoSG	WES	Chen et al., 2020 [G6]	OncoSG	300
data/lung_nci_2022/	NCI	WGS	Zhang et al., 2021 [G7]	NCI Sherlock-Lung	187
data/luad_broad/	Broad	WES, WGS, or both	Imielinski et al., 2012 [G8]	Broad Institute	180
data/yale_luad/	Yale	WES	Kadara et al., 2017 [G9]	Yale University	108
data/luad_cptac_2020/	CPTAC	WES	Gillette et al., 2020 [G10]	CPTAC	108
data/nsclc_tracerx_2017/	TracerX	WES	Jamal-Hanjani et al., 2017 [G11]	TRACERx	61
data/luad_mskcc_2015/	MSK2015	WES	Rizvi et al., 2015 [G12]	MSKCC	10
data/luad_tsp/	TSP	Targeted	Ding et al., 2008 [G13]	Tumor Sequencing Project	— ¹

¹ TSP is merged into integrated_data/ (Source == "TSP") but excluded from the 9,230-sample analytic cohort above (insufficient smoking-history annotation)

References: [G1] AACR Project GENIE Consortium. Cancer Discov. 2017;7(8):818-831. [G2] Hartmaier RJ et al. Cancer Res. 2017;77(9):2464-2475. [G3] Jordan EJ et al. Cancer Discov. 2017;7(6):596-609. [G4] Rizvi H et al. J Clin Oncol. 2018;36(7):633-641. [G5] Cancer Genome Atlas Research Network. Nature. 2014;511(7511):543-550. [G6] Chen J et al. Nat Genet. 2020;52(2):177-186. [G7] Zhang T et al. Nat Genet. 2021;53(9):1348-1359. [G8] Imielinski M et al. Cell. 2012;150(6):1107-1120. [G9] Kadara H et al. Ann Oncol. 2017;28(1):75-82. [G10] Gillette MA et al. Cell. 2020;182(1):200-225.e35. [G11] Jamal-Hanjani M et al. N Engl J Med. 2017;376(22):2109-2121. [G12] Rizvi NA et al. Science. 2015;348(6230):124-128. [G13] Ding L et al. Nature. 2008;455(7216):1069-1075.

1.2 Supporting reference files (data/)

File(s)	Format	Contents
bed_files/*.bed	BED	Capture intervals for the targeted panels (TSP, MSK-IMPACT 341/410/468, FoundationOne); used by cancereffectsizeR for coverage-based mutation-rate denominators.
gene_panels/foundation_one.txt, tsp.txt, msk341.txt, msk410.txt, msk468.txt	gene list	Hugo Symbols in each named panel.
gene_panels/all_panel_genes.txt, all_panel_samples.txt	text list	Union of genes/samples covered by any panel.
gene_panels/genie_panel_genes.txt, genie_panels_used.txt	CSV	Panel gene content and per-sample panel assignment for GENIE.
gene_panels/msk2017_panels_used.txt, msk2018_panels_used.txt	text	Per-sample panel assignment, MSK 2017/2018 cohorts.
genie_9/genomic_information.txt	TSV	Captured intervals per GENIE panel (SEQ_ASSAY_ID).
genie_9/data_clinical_sample.txt	cBioPortal format	GENIE sample-level metadata.
nsclc_tracerx_2017/case_lists/*.txt	cBioPortal case-list	Sample IDs per TRACERx subset ("all", "sequenced").
hg38ToHg19.over.chain	UCSC liftOver chain	GRCh38→hg19 conversion for TCGA/FM-AD (the only two sources on GRCh38).
genes_list.txt	gene list	Driver genes analyzed in the manuscript.

1.3 Integrated / merged datasets (integrated_data/)

Built by harmonizing the sources into one sample table and one mutation table; direct inputs to code/01_CES_variants_calculation_and_plot.R. All files are comma-separated; the unlabeled first column is a row index from the merge and can be ignored.

merged_luad_clinical.txt — one row per sample.

Column	Meaning
Sample ID	Unique sample identifier, harmonized across cohorts.
Smoker	True/False/blank (ever-smoker/never-smoker/not reported).
Stage	Tumor stage as reported by the source cohort; not harmonized across studies.
Progression Free Survival (months)	PFS follow-up, numeric.
Treatment	Treatment info as reported (available for a subset of cohorts).
Overall Survival (months)	OS follow-up, numeric.
Vital Status	Coded per source cohort (commonly cBioPortal-style: NED, AWD, DOD, DUK; -M suffix = with metastasis).
Patient ID	Unique patient identifier (one patient may have multiple samples).
Overall Survival (Months)	Second OS column retained from the merge; equivalent to Overall Survival (months) where both are populated.

merged_luad_maf.txt — one row per somatic variant.

Column	Meaning
Sample ID	Links to merged_luad_clinical.txt.
Chromosome	No chr prefix. GRCh37/hg19 for all sources except TCGA/FM-AD (GRCh38, lifted over in code/01_...R via data/hg38ToHg19.over.chain).
Start_Position	1-based coordinate (same build note as above).
Mutation	chr:pos Ref>Alt.
Reference_Allele, Tumor_Seq_Allele2	Reference / mutant allele.
Source	Cohort code (see Table 1.1).
Variant_Classification	MAF-standard consequence category.
Panel	Assay/panel ID (blank for exome/genome-wide sequencing).

merged_final.txt — merged_luad_maf.txt joined with the clinical columns from merged_luad_clinical.txt by Sample ID.

merged_luad_maf_TransvertToGene.csv — variant-to-gene annotation from cancereffectsizeR's MAF-loading step.

Column	Meaning
Unique_Patient_Identifier	Sample ID.
Chromosome, Start_Position, Reference_Allele, Tumor_Allele	Post-liftover coordinates/alleles (hg19).
variant_type	e.g. snv.
variant_id	chr:pos_Ref>Alt.
genes, top_gene	Overlapping gene(s); primary gene assigned.
top_consequence	Gene_AAchange (e.g. OR8B3_R292S).
prelift_chr, prelift_start	Pre-liftover coordinates (TCGA/FM-AD only).
liftover_strand_flip	TRUE if liftover required a strand flip.

sample_sequencingType_source_info.csv — one row per sample.

Column	Meaning
Unique_Patient_Identifier	Sample ID.
coverage	genome / exome / targeted.
covered_regions	Covered-region reference set for mutation-rate correction.
sig_analysis_grp	Internal batch index for mutational-signature analysis.
maf_source	Fine-grained assay/panel ID.

1.4 Mouse GEMM comparative data (data/*_final.csv)

Tumor-burden measurements from a GEMM system, digitized from a published figure with WebPlotDigitizer, used for the human-vs-mouse "Comparative framework" analysis in code/.

Source: Blair LM, Juan JM, Sebastian L, et al. Oncogenic context shapes the fitness landscape of tumor suppression. Nat Commun. 2023;14:6422

1.5 Precomputed R objects (code/R_data/)

Checkpoints so downstream steps don't require recomputing from raw MAF data: load_maf_cesa_WES_TGS_WGS.rds (loaded/QC-filtered CESAnalysis, all cohorts), cesa_smoking.rds / cesa_nonsmoking.rds (post effect-size estimation, by smoking status), epistasis_variant_TSG_output.Rdata (output of 02_epistasis_variant_TSG.R), gencode.v38lift37.basic.annotation.gtf.Rdata (pre-parsed GENCODE v38lift37/GRCh37 gene annotation).

2. Code

Installation

Tested on macOS, R 4.3.0.

install.packages(c('ggplot2', 'data.table', 'dplyr', 'rtracklayer', 'stringr', 'ggpubr', 'patchwork'))
remotes::install_github("Townsend-Lab-Yale/cancereffectsizeR@v2.10.2", dependencies = TRUE, force = TRUE)
remotes::install_github("Townsend-Lab-Yale/ces.refset.hg19@*release", dependencies = TRUE, force = TRUE)

Workflow

Run in order from within code/ (scripts use relative paths ../data/, ../integrated_data/):

1. 01_CES_variants_calculation_and_plot.R — loads/QC-filters MAF data for all cohorts, lifts over TCGA/FM-AD to hg19, estimates cancer effect sizes (smokers vs. never-smokers), compares to mouse GEMM tumor burden, and generates figures.
2. 02_epistasis_variant_TSG.R — tests variant × TSG epistasis and generates the human-vs-mouse epistasis figures.
3. 03_mouseStats_humanSCC_stat.R — mouse tumor-burden summary statistics by genotype, written to code/SuppleStatistics/.

3. License

• Code (code/*.R): GNU GPL v3.0.
• Data: third-party files (Section 1.1) retain their original source's license/terms (cBioPortal, NCI GDC, or AACR Project GENIE).