Parseq

Parseq estimates local transcription levels and identifies transcript borders from RNA-Seq read counts using a state-space statistical model.

Key Features:

• State-space modeling: Models local transcription levels with a state-space framework that captures both abrupt shifts and gradual drifts in transcription activity.
• Emission model: Implements an emission model that accounts for read count variance, short-range autocorrelation, and the fraction of positions with zero counts within transcripts.
• Estimation via Particle Gibbs: Uses a particle Gibbs algorithm for posterior estimation of latent transcription states.
• Microbial genome efficiency: Computational approach is described as efficient for microbial genomes and large-scale datasets.
• Empirical validation: Demonstrated accurate reconstruction of transcription landscapes in comparative analyses of synthetic and real microbial datasets.

Scientific Applications:

• Microbial Genomics: Analysis of RNA-Seq data from bacteria and other microorganisms to resolve transcriptional structure and boundaries.
• Transcription Research: Investigation of gene expression regulation and transcriptional dynamics at high resolution across conditions or developmental stages.

Methodology:

Inputs are RNA-Seq read counts; transcription levels are modeled with a state-space framework capturing abrupt changes and gradual trends; an emission model captures variance, autocorrelation, and zero-count positions; estimation is performed via a particle Gibbs algorithm.