Pancreatic Cancer Diagnostic Model with 20-miRNA Signature and 801 Samples

801 samples from public datasets were used to develop an interpretable diagnostic model for pancreatic cancer. The model, based on a 20-microRNA signature, achieved cross-validation AUCs up to 0.87 and was validated across independent RNA-seq and qRT-PCR cohorts. Author Yanfei Zhu published the dataset on figshare under a CC-BY-4.0 license, last updated in May 2026.

Use Cases

• Validate a 20-miRNA signature for pancreatic cancer diagnosis based on the described model.
• Benchmark new diagnostic models against the reported AUC performance across RNA-seq and qRT-PCR platforms.
• Investigate the role of specific miRNAs like miR-6875-5p and miR-196a-5p in cancer pathways based on SHAP analysis findings.
• Assess model interpretability frameworks using the SHAP analysis methodology described.

Strengths

• Model was externally validated across multiple independent cohorts totaling 801 samples.
• Achieved a cross-validation AUC of 0.87 with sensitivity of 84.7% and specificity of 83.1%.
• Includes structured validation using serum-based data (GSE128508; n=30) in addition to tissue cohorts.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is based on retrospective datasets; the authors note a need for prospective evaluation.

Provenance

Source

Publicly available datasets (GSE59856, GSE85589, TCGA-PAAD, GTEx, GSE128508) aggregated by author Yanfei Zhu.

Collection Method

Feature selection and Random Forest model development conducted on a training cohort, with independent external validation.

Time Range

null

Freshness

Last updated 2026-05-04 17:37:36; freshness should be verified.

Geography

null