Transcriptional landscape estimation from tiling array data using a model of signal shift and drift

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• High-density oligonucleotide tiling arrays offer genome-wide transcription analysis but are limited by simple statistical models.
• Current models, like the piecewise constant Gaussian model, struggle with determining the optimal number of breakpoints.
• Analyzing complex transcriptional landscapes requires more sophisticated statistical approaches.

Purpose of the Study:

• To develop a new probabilistic methodology for analyzing high-density oligonucleotide tiling array data.
• To overcome limitations of existing statistical models in transcription profiling.
• To provide a more comprehensive understanding of transcriptional dynamics.

Main Methods:

• A hidden Markov model (HMM) framework was developed for signal segmentation.
• The HMM probabilistically embeds signal segmentation, addressing the challenge of fixed breakpoint selection.
• The model incorporates signal drift and covariates for enhanced accuracy.

Main Results:

• The hidden Markov model provides a probabilistic segmentation of continuous signals, improving transcription profile analysis.
• This approach retrieves more information than unique segmentation by accessing the full probability distribution.
• The methodology successfully accounts for subtle effects like signal drift and covariates.
• Demonstrated relevance on a Bacillus subtilis dataset.

Conclusions:

• The developed hidden Markov model offers a powerful and flexible framework for high-density oligonucleotide tiling array data analysis.
• This probabilistic approach enhances the description of transcriptional complexity and dynamics.
• The software is available under the GPL, promoting accessibility and further research.