OutSingle

OutSingle detects outliers in RNA-Seq gene expression (GE) data to identify biologically relevant expression anomalies while controlling for confounders.

Key Features:

• Outlier Detection Using Singular Value Decomposition (SVD): Uses singular value decomposition (SVD) to identify genes with expression levels that deviate significantly from expected patterns in RNA-Seq GE data.
• Optimal Hard Threshold (OHT) Method: Applies the optimal hard threshold (OHT) method, based on SVD, to distinguish signal from noise for confounder control.
• Log-Normal Count Modeling: Adopts a log-normal model for read counts instead of inferring negative binomial distribution (NBD) parameters, reducing computational demands.
• Injection of Artificial Outliers: Injects artificial outliers into RNA-Seq GE data and can mask injected outliers with confounders for simulation and robustness testing.

Scientific Applications:

• Mendelian disorder gene identification: Detects expression outliers to aid identification of genes involved in Mendelian disorders.
• High-throughput genomic studies: Scales to large RNA-Seq datasets for genome-wide outlier detection in high-throughput studies.
• Simulation and pipeline validation: Simulates outliers to test and refine bioinformatics pipelines and assess robustness to confounders.

Methodology:

Combines singular value decomposition (SVD) with the optimal hard threshold (OHT) method and employs a log-normal model for count data.