MBCdeg

MBCdeg applies model-based gene clustering to RNA-seq data to integrate clustering with differential expression analysis and improve detection and classification of differentially expressed genes.

Key Features:

• Model-based clustering (MBCluster.Seq): Implements a model-based gene clustering algorithm via the MBCluster.Seq R package and analyzes all genes in the dataset rather than only prefiltered DEGs.
• Posterior probability assignment: Uses posterior probabilities to assign genes to clusters characterized by non-DEG or DEG expression patterns and to produce an overall gene ranking.
• Comparative performance: Demonstrated, on simulated and real RNA-seq data, improved detection of low-abundance DEGs compared with edgeR, DESeq2, and TCC when P_DEG is below 50%.
• Normalization compatibility (DEGES): Can incorporate the DEGES normalization method to improve stability and consistency of DEG identification when P_DEG is relatively low.
• Expression-pattern classification: Classifies genes by expression patterns across conditions, supporting interpretation in complex designs such as time-course and multi-group studies.

Scientific Applications:

• Time-course experiments: Tracking dynamic gene expression patterns over time and grouping genes by temporal expression profiles.
• Multi-group comparisons: Distinguishing differential expression patterns across multiple experimental groups by cluster-based classification.
• Low-proportion DEG studies: Detecting and ranking DEGs in datasets with a low proportion of differentially expressed genes (low P_DEG).

Methodology:

Uses the MBCluster.Seq R package for model-based clustering on all genes, assigns genes to clusters using posterior probabilities, can apply DEGES normalization, and was evaluated on simulated and real RNA-seq datasets against edgeR, DESeq2, and TCC.