MAMnet: detecting and genotyping deletions and insertions based on long reads and a deep learning approach

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Structural variations (SVs), including deletions and insertions, are crucial to human genetic diversity and linked to various genetic diseases.
• Accurate detection and genotyping of SVs are essential for advancing genetic disease research.
• Long-read sequencing technologies have enhanced SV detection, but challenges remain in achieving optimal accuracy and scalability.

Purpose of the Study:

• To introduce MAMnet, a novel, fast, and scalable method for detecting and genotyping structural variations using long-read sequencing data.
• To leverage a deep neural network, combining convolutional and long short-term memory networks, for sensitive SV detection.

Main Methods:

• Developed MAMnet, a deep learning framework integrating Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks.
• Utilized a novel prediction strategy within the deep neural network for SV detection.
• Evaluated MAMnet on real-world long-read sequencing datasets.

Main Results:

• MAMnet demonstrated superior performance compared to established SV detection tools like Sniffles, SVIM, cuteSV, and PBSV.
• The method achieved higher F1 scores, indicating improved accuracy in SV detection and genotyping.
• MAMnet exhibited enhanced scalability, making it suitable for large-scale genomic analyses.

Conclusions:

• MAMnet offers a significant advancement in the accurate and efficient detection and genotyping of structural variations from long-read sequencing data.
• The proposed deep learning approach provides a robust solution for challenges in SV analysis, benefiting genetic disease research.
• MAMnet's performance and scalability make it a valuable tool for the genomics community.