Unraveling overlapping deletions by agglomerative clustering

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Structural variations, like deletions, are crucial in cancer development.
• Next-Generation Sequencing (NGS) enables detection of these variations.
• Existing tools struggle with accurately calling overlapping deletions.

Purpose of the Study:

• To develop a method for predicting possibly overlapping deletions from short-read paired-end data.
• To address limitations of current tools in handling complex deletion signals.

Main Methods:

• A novel approach using agglomerative clustering to group paired-end mapping data.
• The method iteratively merges mappings based on deletion location and size similarity.
• No assumptions are made regarding data composition (e.g., sample number, heterogeneity).

Main Results:

• The developed method successfully predicts putatively overlapping deletions.
• Erroneous mappings were identified as singleton clusters.
• High accuracy was achieved in distinguishing overlapping from single deletions compared to other methods.

Conclusions:

• Agglomerative clustering is effective for predicting deletions, including overlapping ones.
• The method provides accurate predictions on real and simulated cancer genomic data.
• This approach enhances the analysis of structural variations in cancer research.