Review papers with raw data transparency

Visual first — Key quantitative panoramas from PCAWG (reproduced)

Visual takeaways (figures source)

• PCAWG generated a consensus callset across ~2,600 tumours: ~43.8M SNVs, 2.4M indels, ~288k structural variants, and numerous retrotranspositions and mtDNA mutations — raw counts reproduced above from the paper PCAWG counts and methods
• Chromothripsis, chromoplexy and kataegis are common clustered processes; chromothripsis detected in ~22% of samples and frequently early in evolution; kataegis enriched for APOBEC signatures — see companion analyses Chromothripsis companion paper

Critical scientific evaluation (evidence-weighted)

Strengths — reproducibility, scale, methods

1. Uniform re-alignment and multi-pipeline consensus calling across thousands of whole genomes; Dockerized pipelines + cloud distribution improved reproducibility and enabled public release of workflows and data (Pipelines and validation).
2. Large sample size across 38 tumour types enabled cross-tumour discovery (mutational signatures, SV classes, telomere mechanisms) and robust companion papers (mutational signatures, noncoding drivers, mitochondrial genomes) that add mechanistic layers (Mutational signatures companion).

Key limitations and blind spots (evidence-cited)

1. Indel calling sensitivity and interpretation remain limited using short reads: paper reports indel sensitivity ~40–60% (consensus ~60%) with precision variable; short-read WGS under-detects complex indels and many SV classes; long-read sequencing is required to fully resolve complex SVs and templated insertions (Indel/SV limitations).
2. Non-coding driver discovery remains underpowered: PCAWG and companion noncoding study found relatively few recurrent noncoding drivers (TERT promoter dominates), and the authors caution that many noncoding signals are weak or confounded by local processes or mapping artefacts; larger cohorts and better functional annotation are required (Non-coding drivers companion).
3. Ancestry and tumour-type sampling bias: 77% European ancestry in cohort; some tumour types have small sample sizes, reducing power for subtype-specific discovery and increasing risk of missing population-specific drivers (paper methods and sample tables). See data availability and cohort description (Cohort composition and biases).
4. Clinical translation gap: while the dataset is a vital resource, the heterogeneity and multiple low-frequency events mean clinical predictors demand far larger knowledge-banks with curated outcomes (authors discuss ICGC-ARGO vision). The paper is explicit: precision medicine requires tens of thousands of patients to build robust predictors (Precision-medicine scale needs).

Methodological critique & recommendations

• Variant calling: consensus merging (two-or-more callers for SNVs; logistic stacking for indels) is prudent — but the field now benefits from long-read WGS and graph-based references; reanalysis of PCAWG BAMs with long reads (where available) or hybrid assembly could recover missing indels/SVs and refine driver catalogs.
• Noncoding discovery: incorporate high-resolution regulatory maps (ATAC/ChIP-seq from matched tissues), integrate eQTL/eGene maps and single-cell profiles, and perform CRISPR perturbation screens of top candidate noncoding loci to validate functionality.
• Population representation: future expansions should prioritize underrepresented ancestries and tumor types to discover population-specific drivers and L1 source-element activity differences documented by PCAWG.
• Functional validation: PCAWG properly treats many noncoding and structural findings as hypotheses; systematic functional pipelines (MPRA, CRISPRi/CRISPRa, enhancer-swap assays) should be applied to highest-confidence candidates.

Where PCAWG changed practice and open resources it provided

• Open, dockerized pipelines on Dockstore; harmonized data portal (ICGC DCC) and Synapse mirrors greatly improve reproducibility and reuse (Data + pipelines availability).
• Companion papers (signatures, SVs, noncoding drivers, telomeres, mtDNA) constitute a modular, multi-omic resource enabling targeted follow-ups (citations: mutational signatures, chromothripsis, noncoding drivers, telomeres, mtDNA) (Signatures companion)

Confidence, reproducibility, and data access

Confidence in main PCAWG results (consensus SNV/SV landscapes, signatures, large-scale patterns) is high because of validation experiments, multiple pipelines, and companion replication studies; reproducibility is strengthened by Dockstore images and ICGC portals. However, claims about rare noncoding drivers or some SV categories should be treated as provisional pending longer reads, larger cohorts, and experimental validation (Validation and caution statements).

Suggested immediate, practical follow-ups for a researcher

1. Re-run PCAWG consensus VCFs through long-read-aware SV integrators (if long-read data are available for matched samples) or perform local hybrid assembly for candidate regions (e.g., TERT, CCND1 amplifications, templated insertions) to validate complex SVs.
2. Take the top noncoding candidate list from the PCAWG noncoding companion and design high-throughput MPRA or CRISPRi screens in relevant cell types to test allelic regulatory activity (prioritize TERT-neighbor hits and 3'UTR candidates with expression associations).
3. Use PCAWG germline–somatic association results to prioritize patients for functional follow-up: e.g., BRCA1-associated templated insertions and MBD4 germline PTVs increasing CpG mutagenesis (Germline–somatic associations).

Concise conclusion

PCAWG is a milestone: it created a reproducible, richly annotated pan-cancer whole‑genome resource and catalysed multiple companion discoveries (signatures, SV patterns, telomere mechanisms, mitochondrial variation, noncoding candidate drivers). It also highlights persistent limits (short-read indel/SV sensitivity, power to detect noncoding drivers, ancestry/tumour sampling gaps). Use PCAWG as a foundation, but combine it with long reads, functional assays, and larger, more diverse cohorts to close the remaining discovery gaps.

Key citations used (primary PCAWG + high-value companions)

PCAWG primary paper

Noncoding drivers paper

Mutational signatures resource

Practical links & actions

Download PCAWG data and pipelines to reproduce analyses: ICGC Data Portal and Dockstore (links in paper). Useful immediate queries you can run next with BGPT:

How I scored the paper (concise metrics)

• Novelty: 8 — first integrated pan-cancer WGS resource at this scale with many companion mechanistic papers.
• Quality: 9 — extensive validation, benchmarked pipelines, open code/data, strong statistical practice.
• Generality: 7 — broad across tumor types but limited by ancestry/tumour sampling imbalances.
• Usefulness: 9 — major community resource for discovery and methods benchmarking.
• Reproducibility: 9 — Dockerized pipelines, Dockstore, ICGC portals; residual reproducibility limits for indels/SVs due to short-read tech.
• Explanatory depth: 8 — deep descriptive and mechanistic companion studies but causal inference limited for some noncoding hits.

If you want, I can: (A) run a bespoke re-analysis of PCAWG consensus VCFs to re-evaluate indel burden in promoters and the TERT hotspot coverage, (B) build an MPRA prioritization list of top noncoding candidates for functional testing, or (C) create an analysis pipeline recommendation (snakemake/nf-core) to combine short- and long-read SV calls. Click below to start an AI Scientist agent to run live bioinformatics workflows.

Author Reviews (quick links)

Open bespoke author-review queries (one-click) for the first/senior authors — use these to request targeted author-level critiques or follow-ups:

If you'd like a deeper reanalysis (e.g., recalc power curves for noncoding drivers, produce a ranked MPRA candidate list, or re-derive signature attributions for a tumour type), press the Run AI Scientist Analysis button above and upload any VCFs/BAMs you have — the agent will iteratively run the workflows and return reproducible results.