Cyber-T

Cyber-T performs Bayesian-regularized differential expression analysis of high-throughput biological data, including microarrays, next-generation sequencing (RNA-seq), protein arrays, and quantitative mass spectrometry, using regularized t-tests to estimate variances.

Key Features:

• Bayesian-regularized variance estimation: Implements the Baldi and Long Bayesian framework to stabilize variance estimates for each probe or measurement.
• Regularized t-test framework: Computes regularized t-statistics by combining empirical measurements with priors derived from pooled neighborhoods.
• Multi-platform support: Applies to microarrays, next-generation sequencing (RNA-seq), protein arrays, and quantitative mass spectrometry datasets.
• Neighborhood-based priors: Derives priors from pooled data within similar neighborhoods to mitigate effects of low replication and platform-specific biases.
• Normalization options: Includes logarithmic transformations and Variance Stabilizing Normalization (VSN) for preprocessing.
• Statistical tests: Supports two-sample t-tests and one-way analysis of variance (ANOVA).
• Multiple testing correction: Provides standard frequentist correction methods and a probabilistic mixture model for multiple-testing control and false discovery rate estimation.
• Diagnostic visualization: Generates diagnostic plots for visual assessment of variance estimates and test results.

Scientific Applications:

• Differential expression (microarrays): Identification of differentially expressed genes from DNA microarray experiments using regularized variance estimates.
• Differential expression (RNA-seq): Analysis of next-generation sequencing (RNA-seq) count data for expression changes using the regularized t-test framework.
• Proteomics analyses: Differential analysis of protein arrays and quantitative mass spectrometry datasets to detect altered protein abundances.
• Low-replication studies and bias mitigation: Stabilization of variance estimates and control of false discoveries in experiments with few replicates or platform-specific biases.

Methodology:

Implements Bayesian regularization (Baldi and Long) to compute regularized variances within a t-test framework, uses pooled-neighborhood priors, offers logarithmic and VSN normalization, supports two-sample t-tests and one-way ANOVA, and applies frequentist multiple-testing corrections or a probabilistic mixture model with diagnostic plots.