PLNseq

PLNseq performs differential expression analysis on RNA-seq read count data from correlated samples by modeling inter-sample correlation to detect differentially expressed genes in matched or repeated-measures experimental designs.

Key Features:

• Modeling Correlation: Employs a multivariate Poisson lognormal distribution with Gaussian random effects to capture correlation between read counts across samples.
• Likelihood-Based Inference: Uses likelihood methods for inference about differential expression.
• Three-Stage Numerical Algorithm: Implements a three-stage numerical algorithm to estimate unknown parameters efficiently.
• False Discovery Rate Control: Demonstrates superior control of false discovery rates (FDRs) compared to edgeR and DESeq2 when gene-specific correlations vary but can be estimated.
• Enhanced DE Testing: Directly evaluates correlation to enable a novel and more powerful test for differential expression.

Scientific Applications:

• Simulated Data Validation: Validated on simulated data to assess parameter estimation accuracy, robustness, and analytical power.
• Lung Cancer Study: Applied to a lung cancer RNA-seq dataset to illustrate practical performance and biological insight.

Methodology:

Modeling via a multivariate Poisson lognormal distribution with Gaussian random effects; inference using likelihood methods; parameter estimation by a three-stage numerical algorithm; direct evaluation of correlation for differential expression testing and comparative FDR assessment against edgeR and DESeq2.