IBS-CNN: a novel computer vision-based kinship classification model for high-density SNP microarray.

Area of Science:

• Genetics
• Forensic Science
• Bioinformatics

Background:

• Forensic genealogy faces challenges in diverse populations like Chinese.
• Existing kinship classification methods require adaptation for specific genetic data.

Purpose of the Study:

• To develop and evaluate novel machine learning (ML) and deep learning (DL) strategies for kinship classification in Chinese populations.
• To assess the performance of traditional methods against new ML and DL approaches.

Main Methods:

• Utilized Infinium Asian Screening Array (ASA) and 1000 Genomes Project data for simulations.
• Developed a multi-feature ML model and an IBS-CNN model integrating computer vision with identity-by-state (IBS) theory.
• Evaluated shared identity-by-descent (IBD) segment analysis, likelihood ratio (LR), and kinship coefficient methods.

Main Results:

• The IBS-CNN model demonstrated superior performance and robustness compared to traditional methods and ML models.
• Achieved 94.56% accuracy on real data, processing IBS heatmap images securely and efficiently.
• The IBS-CNN model offers a more secure and storage-efficient data transmission format than raw DNA data.

Conclusions:

• The IBS-CNN model represents a significant advancement in applying forensic genealogy to Chinese populations.
• This deep learning approach provides a robust, accurate, and efficient tool for kinship analysis.
• The developed method enhances data security and storage efficiency in genetic genealogy applications.