A Poisson hierarchical modelling approach to detecting copy number variation in sequence coverage data.

Area of Science:

• Genomics
• Bioinformatics
• Microbiology

Background:

• Next-generation sequencing accelerates the study of copy number variation (CNV) in microbial genomes for public health.
• CNV analysis typically involves read mapping, coverage calculation, and detection of deletions/amplifications.
• Existing CNV detection methods often rely on strict statistical assumptions or require algorithm tuning.

Purpose of the Study:

• To introduce a novel CNV detection methodology using two Poisson hierarchical models (Poisson-Gamma and Poisson-Lognormal).
• To develop a flexible and robust approach for CNV detection that accommodates diverse data patterns and deviations from the Poisson model.

Main Methods:

• Utilized Poisson-Gamma and Poisson-Lognormal hierarchical models for CNV detection.
• Applied the methodology to sequence coverage data from seven Plasmodium falciparum malaria genomes.
• Validated the approach using 3D7 resequencing data and simulations.

Main Results:

• Empirical coverage distributions in Plasmodium falciparum genomes are asymmetric and overdispersed relative to the Poisson model.
• The proposed methodology demonstrated a low false positive rate.
• Detected known CNV regions (e.g., PfMDR1 amplification, CLAG3.2 deletion) and identified putative novel CNVs.
• Showed higher concordance with array data compared to FREEC and cn.MOPS for 7G8 and GB4 isolates.

Conclusions:

• The new methodology offers enhanced flexibility, robustness, accuracy, and statistical rigor for CNV detection from sequence coverage data.
• This advancement is crucial for accurate genomic analysis in public health microbiology.