From: Comprehensive evaluation of differential gene expression analysis methods for RNA-seq data
Evaluation | Cuffdiff | DESeq | edgeR | limmaVoom | PoissonSeq | baySeq |
---|---|---|---|---|---|---|
Normalization and clustering | All methods performed equally well | |||||
DE detection accuracy measured by AUC at increasing qRT-PCR cutoff | Decreasing | Consistent | Consistent | Decreasing | Increases up to log expression change ≤ 2.0 | Consistent |
Null model type I error | High number of FPs | Low number of FPs | Low number of FPs | Low Number of FPs | Low number of FPs | Low number of FPs |
Signal-to-noise vs P value correlation for genes detected in one condition | Poor | Poor | Poor | Good | Moderate | Good |
Support for multi-factored experiments | No | Yes | Yes | Yes | No | No |
Support DE detection without replicated samples | Yes | Yes | Yes | No | Yes | No |
Detection of differential isoforms | Yes | No | No | No | No | No |
Runtime for experiments with three to five replicates on a 12 dual-core 3.33 GHz, 100 G RAM server | Hours | Minutes | Minutes | Minutes | Seconds | Hours |