Fast Bayesian Inference of Copy Number Variants using Hidden Markov Models with Wavelet Compression

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Genomic copy number variants (CNV) are crucial in genetic variation and disease.
• Accurate detection of CNVs from array comparative genomic hybridization (aCGH) data is computationally intensive.
• Existing methods for CNV detection face challenges in speed and efficiency.

Purpose of the Study:

• To develop a computationally efficient method for detecting genomic copy number variants (CNV).
• To improve the accuracy and speed of Bayesian inference for CNV detection in array CGH data.
• To enable routine diagnostic use and re-analysis of large genomic datasets.

Main Methods:

• Integration of Haar wavelets with Hidden Markov Models (HMMs).
• Application of Forward-Backward Gibbs sampling for Bayesian inference.
• Dynamic and adaptive recomputation of observational blocks for focused analysis.
• Development of an automatic prior for enhanced usability.

Main Results:

• Drastically reduced running times for Bayesian inference compared to standard Gibbs sampling.
• Improved detection of genomic copy number variants (CNV) in array CGH experiments.
• Concentration of computational effort on challenging chromosomal segments.
• Demonstrated feasibility for routine diagnostics and legacy data re-analysis.

Conclusions:

• The Haar wavelet-HMM approach offers a significant advancement in the speed and efficiency of CNV detection.
• The method is suitable for practical diagnostic applications and large-scale genomic data analysis.
• An open-source software implementation (HaMMLET) is available, facilitating adoption and further research.