PDATK

PDATK provides an S4 class-based interface for performing unsupervised subtype discovery, cross-cohort meta-clustering, gene-expression-based classification, and subsequent survival analysis to identify prognostically useful subtypes in pancreatic cancer and beyond.

Key Features:

• Consensus Subtypes in Pancreatic Cancer (CSPC): A novel method for consensus-based meta-clustering across multiple cohorts.
• Pancreatic Cancer Overall Survival Predictor (PCOSP): A novel method for overall-survival prediction based on gene expression data.
• Published Subtype Classifiers: Four published subtype classifiers are included for recreating published results, applying existing classifiers to new data, and benchmarking new methods.
• Published Prognostic Gene Signatures: Three published prognostic gene signatures are included for similar validation and benchmarking purposes.
• Bioconductor Integration: All PDATK classes and methods use existing Bioconductor classes as input, enabling integration with existing Bioconductor datasets including the 21 pancreatic cancer patient cohorts in the MetaGxPancreas data package.
• Subtype Centroids and Prognostic Gene Signatures: Inclusion of these from publications enables classification of novel patient gene expression data for direct clinical application.

Scientific Applications:

• Pancreatic Cancer Subtype Discovery: Identification of meaningful molecular subtypes from gene expression profiles to inform therapeutic strategy in pancreatic ductal adenocarcinoma.
• Cross-Cohort Meta-Clustering: Integration and consensus clustering across multiple patient cohorts to identify robust, reproducible subtypes.
• Gene-Expression-Based Classification: Application of existing and novel classifiers to categorize patient samples into molecular subtypes.
• Survival Analysis: Subsequent analysis to identify prognostically useful subtypes and evaluate classifier performance.
• Classifier Benchmarking: Comparison of new methods against four published subtype classifiers and three published prognostic gene signatures.
• Clinical Translation: Classification of novel patient gene expression data using included subtype centroids and prognostic gene signatures to inform clinical decision making.

Methodology:

PDATK implements an S4 class-based architecture with methods for unsupervised subtype discovery, cross-cohort meta-clustering, gene-expression-based classification, and survival analysis. CSPC performs consensus-based meta-clustering and PCOSP performs overall-survival prediction. The package uses existing Bioconductor classes as input and integrates with the MetaGxPancreas data package containing 21 pancreatic cancer patient cohorts.