Linear: a framework to enable existing software to resolve structural variants in long reads with flexible and...

Area of Science:

• Genomics
• Bioinformatics

Background:

• Structural variant (SV) detection is crucial for long-read sequencing analysis.
• Current alignment-based methods face challenges with forced alignments, model integration, and computational speed.

Purpose of the Study:

• To investigate the feasibility of alignment-free algorithms for long-read SV detection.
• To evaluate if alignment-free approaches offer advantages over existing methods.

Main Methods:

• Implementation of the Linear framework, integrating alignment-free algorithms for SV detection.
• Development of a generative model for long-read SV detection within Linear.
• Ensuring compatibility of alignment-free outputs with existing bioinformatics software.

Main Results:

• Linear demonstrates superior sensitivity and flexibility compared to alignment-based pipelines.
• The framework achieves orders of magnitude faster computational efficiency.
• Successful integration of alignment-free SV detection into standard bioinformatics workflows.

Conclusions:

• Alignment-free approaches are feasible and advantageous for long-read SV detection.
• The Linear framework provides a sensitive, flexible, and computationally efficient solution.
• Linear facilitates the integration of novel SV detection models and improves compatibility with existing tools.