STAN

STAN segments genomes using bidirectional hidden Markov models (HMMs) to infer directed genomic states from strand-specific and non-strand-specific occupancy profiles for the characterization of strand-directed processes such as DNA replication, transcription, and repair.

Key Features:

• Bidirectional Hidden Markov Models (HMMs): Implements bidirectional HMMs that infer directed genomic states from occupancy profiles de novo.
• Integration of Strand-specific and Non-strand-specific Data: Integrates strand-specific signals (e.g., RNA expression) with non-strand-specific signals (e.g., ChIP-seq) for joint genome segmentation.
• Multiple Probability Distributions: Employs various probability distributions within the HMM framework to model diverse genomic datasets.
• De Novo Genome Annotation: Learns and annotates the genome into a user-specified number of genomic states without prior state definitions.
• Directional State Inference from Occupancy Profiles: Infers the directionality of protein complexes moving along the genome to characterize strand-specific biological processes.

Scientific Applications:

• Yeast transcription analysis: Applied to RNA polymerase II-associated factors in yeast to identify 32 new transcribed loci and to reveal gene-specific variations and a regulated initiation–elongation transition.
• Human chromatin state patterns: Applied to human T cells to indicate directed chromatin state patterns at transcribed regions in contrast to repressed areas, informing transcriptional regulation and chromatin dynamics.
• Characterization of strand-specific processes: Used to characterize strand-directed processes including DNA replication, transcription, and repair by assigning directional genomic states.

Methodology:

Applies bidirectional HMMs to occupancy profiles, accommodates both strand-specific and non-strand-specific signals, uses multiple probability distributions within the HMM, and performs de novo learning of genomic states.