ConsensuSV-ONT - A modern method for accurate structural variant calling.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Advancements in sequencing technologies drive the need for robust structural variant detection tools.
• Existing tools for Oxford Nanopore (ONT) long-read sequencing are limited, posing challenges for optimal tool selection.
• The integration of machine learning, particularly deep learning, offers new avenues for improving variant analysis.

Purpose of the Study:

• To develop a novel, automated algorithm, ConsensuSV-ONT, for high-quality structural variant detection in ONT long-read data.
• To consolidate and enhance existing structural variant calling methods through a consensus-based approach.
• To provide an accessible and user-friendly tool for researchers working with ONT sequencing data.

Main Methods:

• Integration of six state-of-the-art structural variant callers for long-read sequencing.
• Application of a convolutional neural network (CNN) for filtering and quality control of identified variants.
• Development of a Docker image and Nextflow pipeline for efficient, parallelized data processing.

Main Results:

• The ConsensuSV-ONT algorithm successfully identifies a set of high-quality, reliable structural variants.
• The method leverages deep learning for enhanced accuracy in variant filtering.
• A complete, ready-to-use runtime environment is provided, facilitating broader adoption.

Conclusions:

• ConsensuSV-ONT addresses the need for improved structural variant detection in Oxford Nanopore sequencing.
• The algorithm offers a robust and automated solution, combining multiple callers and deep learning.
• The tool is designed for ease of use, benefiting both bioinformaticians and researchers in genomics.