Performance testing for the sensitivity and resolution of low-pass WGS for small CNV detection

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Low-pass genome sequencing (LP GS) is a standard method for detecting copy number variations (CNVs).
• Window selection is a critical algorithmic parameter in LP GS that impacts performance, particularly for small CNVs.
• Previous research has limitedly explored the influence of window selection on small CNV detection.

Purpose of the Study:

• To evaluate the impact of sliding window parameters on the true positive rate, interpretation workload, and resolution of LP GS for CNV detection.
• To compare the performance of different sliding window algorithms using simulated and clinical datasets.
• To provide recommendations for optimal window selection and sequencing depth for detecting small CNVs.

Main Methods:

• Simulated 40 samples with 19 predefined CNVs across various read depths.
• Analyzed 57 clinical cases with existing CMA results (27 positive, 30 negative).
• Evaluated sliding window algorithms (10-Kb window/1-Kb increments vs. 50-Kb window/5-Kb increments) for sensitivity, specificity, and resolution.

Main Results:

• A 10-Kb window with 1-Kb increments demonstrated a higher true positive rate, especially for CNVs ≤30 Kb, achieving 100% for 30 Kb deletions.
• This 10-Kb window algorithm showed less variability and higher resolution for small deletions and duplications.
• Clinical case evaluation showed 96.30% sensitivity for the 10-Kb window algorithm versus 85.19% for the 50-Kb window algorithm, with comparable specificity (96.67%).

Conclusions:

• Window selection and sequencing depth are crucial for CNV detection sensitivity and resolution in LP GS, particularly for small CNVs.
• A 10-Kb window with 1-Kb increments and ≥50 M reads is recommended for detecting CNVs ≤30 Kb.
• Clinical laboratories should establish sensitivity and resolution benchmarks for different sliding window and sequencing depth combinations.