BUMHMM

BUMHMM models high-throughput RNA structure probing data to compute per-nucleotide posterior probabilities of nucleotide modification.

Key Features:

• Probabilistic Modeling: Computes per-nucleotide posterior probabilities using a probabilistic model that accounts for biological variability and biases.
• Bias Correction: Empirically corrects for coverage- and sequence-dependent biases observed in structure probing sequencing data.
• Drop-off Rate Utilization: Incorporates a per-nucleotide "drop-off rate" measure into the model as an informative signal.
• Support for Multiple Replicates: Jointly analyzes multiple experimental replicates to improve inference robustness.

Scientific Applications:

• Transcriptome-wide Analysis: Produces per-nucleotide modification probabilities across transcriptomes to enable studies of RNA structure in gene regulation.
• Increased Sensitivity: Identifies modified regions with higher sensitivity compared to existing pipelines.
• Lower Coverage Requirements: Enables confident modification calls at lower sequencing coverage than typically recommended for structural probing experiments.

Methodology:

Uses probabilistic modeling to compute per-nucleotide posterior probabilities from structure probing combined with high-throughput sequencing, incorporates per-nucleotide drop-off rates, applies empirical correction for coverage- and sequence-dependent biases, and performs joint analysis of multiple replicates.