CORNAS

CORNAS models true gene counts from RNA-Seq data using sequencing coverage and Bayesian inference to improve differential gene expression analysis under limited or absent biological replicates and low sequencing coverage.

Key Features:

• Incorporation of sequencing coverage: Uses sequencing coverage information derived from RNA sample concentration as an explicit analysis parameter.
• Bayesian posterior estimation: Applies Bayesian methods to estimate the posterior distribution of true gene counts conditioned on observed counts and coverage.
• Representation of count uncertainty: Models that a single observed gene count can correspond to multiple potential true gene counts with varying probabilities.
• Integration into differential expression testing: Integrates coverage-informed posterior distributions into differential gene expression analysis workflows.
• Designed for low coverage and few replicates: Tailored for experiments with low sequencing coverage and limited or absent biological replicates.
• Performance benchmarking: Simulation studies and experimental validations reported performance comparable to or superior to NOISeq and GFOLD.
• Accounts for normalization assumptions: Explicitly addresses uncertainty in the assumption that adjusted observed counts represent unknown true gene counts.

Scientific Applications:

• Differential expression with limited replicates: Identification of differentially expressed genes from RNA-Seq when biological replicates are scarce or absent.
• Low sequencing coverage analysis: Refinement of gene count interpretation in low-coverage RNA-Seq experiments.
• Benchmarking and validation: Validation and comparison of differential expression results via simulation studies and real experimental data against methods such as NOISeq and GFOLD.

Methodology:

Uses a sequencing coverage parameter derived from RNA sample concentration and Bayesian methods to estimate posterior distributions of true gene counts, which are integrated into differential gene expression testing.