iCOBRA

iCOBRA provides standardized, transparent, and reproducible benchmarking, implemented as an R/Bioconductor package, for methods that perform binary classification, ranking, and continuous target estimation in life-science research.

Key Features:

• R/Bioconductor implementation: Provided as an R/Bioconductor package for integration with R-based analysis workflows.
• Accepted input types: Accepts method outputs such as p-values, scores, ranks, posterior probabilities, and binary decisions together with truth labels.
• Performance summaries: Generates receiver operating characteristic (ROC) curves, precision–recall curves, false discovery rate (FDR) assessments, calibration plots, and other evaluation metrics relevant to ranking and classification tasks.
• Comparative assessment: Performs comparative performance assessment of methods against known ground-truth data.
• Curated benchmarking datasets: Provides a collection of curated benchmarking datasets in standardized formats for reproducible evaluation.
• Ranking and continuous-target evaluation: Supports evaluation workflows for both ranked outputs and continuous target estimation.

Scientific Applications:

• Differential expression analysis: Benchmarking of methods that identify differentially expressed genes or transcripts.
• Differential abundance testing: Evaluation of methods for differential abundance in metagenomics or other count-based data.
• Biomarker discovery: Comparative assessment of biomarker ranking and selection methods.
• Statistical filtering pipelines: Assessment of statistical filtering and feature-selection pipelines that produce ranked or classified feature lists.
• Genomics and transcriptomics: Application to ranking and classification tasks in genomics, transcriptomics, and other data-intensive disciplines.

Methodology:

Accepts method outputs (p-values, scores, ranks, posterior probabilities, or binary decisions) with truth labels and computes ROC curves, precision–recall curves, false discovery rate assessments, calibration plots, and other evaluation metrics; includes curated benchmarking datasets in standardized formats.