Probabilistic models of biological enzymatic polymerization.

Area of Science:

• Bioinformatics
• Molecular Biology
• Parasitology

Background:

• Mitochondrial messenger RNAs (mRNAs) in Trypanosoma brucei undergo untemplated addition of adenine and uracil to their 3' ends.
• Understanding this post-transcriptional modification is crucial for gene expression regulation in this human pathogen.

Purpose of the Study:

• To evaluate probabilistic models, specifically Hidden Markov Models (HMMs), for characterizing and generating untemplated adenine/uridine tail populations in T. brucei mitochondrial mRNAs.
• To identify distinct nucleotide addition states and explore potential novel biological insights.

Main Methods:

• Application of hierarchical Hidden Markov Models (HMMs) to analyze nucleotide sequences of 3' untemplated tails.
• Evaluation of HMMs' generative capabilities for tail length distribution and nucleotide composition.
• Correlation of identified HMM states with experimental data and known biological processes.

Main Results:

• HMMs effectively characterized nucleotide composition aspects of the tail populations, though not tail length distribution.
• Distinct nucleotide addition states were robustly identified, correlating with experimentally verified compositional differences.
• A surprising subclass of tails in ND1 gene transcripts was discovered, challenging current models of sequential enzymatic action.

Conclusions:

• The developed HMMs serve as valuable tools for reflecting known biological states and generating testable hypotheses.
• These models offer a method for correcting sequencing errors in the analyzed data.
• The HMMs provide simple, pedagogical examples of applied bioinformatic Hidden Markov Models due to their binary emissions.