derfinderHelper

derfinderHelper enhances derfinder workflows to detect differentially expressed regions (DERs) in RNA-seq data at base-pair resolution using an annotation-agnostic bump-hunting approach and multicore computation.

Key Features:

• Annotation-Agnostic Approach: Identifies DERs without relying on pre-existing gene annotations or transcript assemblies, enabling detection at single-base resolution.
• Bump-Hunting Algorithm: Implements a computationally efficient bump-hunting method to detect DERs genome-wide, supporting large-scale analyses.
• Flexible Statistical Modeling: Provides a statistical framework that supports multi-group comparisons and time-course study designs.
• Advanced Data Visualization: Offers visualization techniques for expressed region analysis at base resolution to explore expression outside known gene boundaries.
• Multicore/Parallel Processing: Leverages multiple computational cores to expedite derfinder analyses.
• Application to Public Datasets: Has been applied to GTEx and BrainSpan RNA-seq datasets to analyze hundreds of samples at base resolution in R.
• Simulation Studies: Demonstrates in simulations that base-resolution approaches remain effective with incomplete annotations and approach feature-level analysis power when annotations are complete.

Scientific Applications:

• Annotation-agnostic DER discovery: Genome-wide detection of differentially expressed regions and novel transcriptional signals independent of gene models.
• Comparative and time-course studies: Differential expression analyses across multiple groups and time points using flexible statistical models.
• Large-scale RNA-seq analysis: Base-resolution analysis of hundreds of samples, exemplified by GTEx and BrainSpan applications in R.
• Method benchmarking and evaluation: Performance assessment under incomplete annotation via simulation studies to compare base-resolution and feature-level approaches.

Methodology:

Annotation-agnostic base-resolution analysis using derfinder methodology, computationally efficient bump-hunting, flexible statistical modeling for multi-group and time-course designs, expressed-region visualization, multicore processing, and evaluation via simulation studies.