Bayesian Integrative Detection of Structural Variations With False Discovery Rate Control

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long-read sequencing advances enable structural variation (SV) detection in genetic diseases.
• Existing methods for merging SVs from multiple tools lack robust confidence quantification.
• Accurate SV detection is crucial for understanding genetic disease mechanisms.

Purpose of the Study:

• To develop a Bayesian integration model for combining structural variation (SV) calls from diverse tools.
• To introduce a method for false discovery rate (FDR) control and quantitative confidence measures for merged SVs.
• To create a flexible model capable of incorporating SV callers with varying quality score availability.

Main Methods:

• A Bayesian integration model was developed to merge structural variation (SV) calls.
• A novel approach for false discovery rate (FDR) control was implemented.
• The model's performance was evaluated using extensive simulation studies and the HG002 dataset.

Main Results:

• The Bayesian model accurately estimates the false discovery rate (FDR).
• The integrated approach significantly improves the F1 score for structural variation (SV) detection.
• The model demonstrates flexibility in handling SV callers with missing quality scores.

Conclusions:

• The proposed Bayesian integration model offers a reliable method for quantifying confidence in merged structural variation (SV) calls.
• This approach enhances the accuracy of SV detection, aiding in the study of genetic diseases.
• The model provides a valuable tool for researchers utilizing long-read sequencing data.