Method

This page explains how the preserved 2019 skin-cancer PRSweb release was reconstructed from the original HTML pages, the PLoS Genetics manuscript, and the published supplement tables.

Use it when you want to understand which results are true score-evaluation summaries, which pages are preserved PRS-PheWAS validations, and where the current static portal had to carry forward historical gaps.

The underlying resource is described in Fritsche LG, Beesley LJ, VandeHaar P, Peng RB, Salvatore M, Zawistowski M, Gagliano Taliun SA, Das S, LeFaive J, Kaleba EO, Klumpner TT, Moser SE, Blanc VM, Brummett CM, Kheterpal S, Abecasis GR, Gruber SB, Mukherjee B. Exploring various polygenic risk scores for skin cancer in the phenomes of the Michigan genomics initiative and the UK Biobank with a visual catalog: PRSWeb. PLoS Genet. 2019 Jun 13;15(6):e1008202. . The local manuscript PDF, supplement PDF, and supplement workbook remain the primary reconstruction sources.

Method in brief

Use Scores to compare preserved PRS models for the same skin-cancer trait, then open a score detail page for metrics, downloads, and linked PheWAS results.
MGI rows retain manuscript-backed fit summaries, while UKB rows mainly preserve the cohort-specific PheWAS validation pages and matched-count associations.
The historical portal reported quartile-enrichment forests, not the newer top-percentile summaries used by later PRSweb releases, so top-percentile fields remain blank here.
Basal and squamous subtype validation in UKB used the broader phecode 172.2 because the preserved UKB phenome did not expose subtype-specific rows.

Evaluation cohorts

The original study combined two related but distinct outputs: score evaluation in the Michigan Genomics Initiative (MGI) and PRS-PheWAS exploration in both MGI and UK Biobank.

MGI is the only cohort for which the preserved manuscript tables report AUC, Brier score, and other model-level fit summaries
UKB pages in this release preserve the paired PRS-PheWAS validation scans shown in the legacy portal
The static release therefore mixes MGI score summaries with cohort-specific PheWAS artifacts when that is all the 2019 source bundle preserved

Source phenotypes

The preserved portal follows the paper's phecode-based skin-cancer phenotype layout. Overall skin cancer, melanoma, basal cell carcinoma, and squamous cell carcinoma each appear as PRS targets, but the exact validation phenotype can differ by cohort.

MGI pages preserve exact subtype rows such as 172.11, 172.21, and 172.22
UKB melanoma pages preserve subtype-specific melanoma rows, but basal and squamous pages resolve to the broader 172.2 “other non-epithelial cancer of skin” phenotype
The supplement workbook supplies the ICD9/ICD10 mapping tables used by the current phecode pages

Construction strategies

The 2019 portal compared three classes of PRS sources: published GWAS Catalog loci, newer trait-specific GWAS summary statistics, and UK Biobank summary-statistic depth experiments.

GWAS Catalog and Latest GWAS models are preserved for melanoma, basal cell carcinoma, and squamous cell carcinoma
Fixed threshold UKB models preserve the six p-value depth comparisons from Supplement Table D
LDpred rows preserve the best-performing causal-fraction settings highlighted in Supplement Table E
Local score downloads point to the preserved GRCh37 weight files from the original snapshot

Score interpretation

Score detail pages combine the best preserved model metadata from three places: the legacy HTML page, the manuscript tables, and the supplement tables. That means different metrics have different provenance.

P-value, matched cases/controls, and quartile odds ratios come from the preserved legacy page payload for the corresponding cohort page
AUC and Brier score for MGI rows are backfilled from Table 2 or Supplement Tables D/E
Nagelkerke pseudo-R² was only reported for the LDpred comparisons in Supplement Table E and remains blank for other historical rows
Top 1%, 2%, 5%, 10%, and 25% fields are blank because that newer summary format was not present in the 2019 portal source

PheWAS interpretation

Every preserved model keeps both the all-subject scan and the exclusion scan from the corresponding legacy page. These are the most complete cohort-specific artifacts in the release.

MGI pages cover 1,578 phenotypes in the legacy portal snapshot
UKB pages cover 1,366 phenotypes in the legacy portal snapshot
All-subject and exclusion scans should be read together, especially for dermatologic follow-up traits such as actinic keratosis and sebaceous-gland phenotypes
When a UKB subtype page uses phecode 172.2, the score page description now calls that out explicitly

Release provenance

This instance is intentionally read-only and versioned as a reconstructed historical release. The release folder is built from a 2019 HTML snapshot plus supplement tables, then normalized into the current static-site contract so the routes, downloads, and PheWAS views remain reproducible.

The release preserves the original MGI and UKB page split rather than collapsing everything into one synthetic summary. Where the legacy bundle lacked a field used by newer portals, the static release leaves that field blank instead of inventing values.

The preserved portal snapshot is dated 2019-06-14 and the underlying manuscript was published on 2019-06-13.

Manuscript acknowledgments

The authors acknowledge the Michigan Genomics Initiative participants, Precision Health at the University of Michigan, the University of Michigan Medical School Data Office for Clinical and Translational Research, the Central Biorepository, the Advanced Genomics Core, and the Center for Statistical Genetics for genotype curation, imputation, and management support.

The original study also used UK Biobank data and phenome validation under application number 24460. Additional source-specific acknowledgments remain in the manuscript and supplement.

Skin Cancer PRSweb