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

Authors

Fanzhang Lei, Qinglin Liang, Xi Yuan, Qinglin Liu, Yu Zhang, Tong Xie, Le Wang, Bofeng Zhu

Published in

International journal of legal medicine. Jun 19, 2026. Epub Jun 19, 2026.

Abstract

This study addresses multiple challenges in applying forensic genealogy to Chinese populations by exploring novel kinship classification strategies based on machine learning (ML) and deep learning (DL). Utilizing Infinium Asian Screening Array (ASA) microarray integrated with Han Chinese reference data from the 1000 Genomes Project, we simulated 70,000 pairs of first- to fifth-degree relatives and unrelated individuals while experimentally validating data from Han Chinese volunteers from 19 pedigrees, including corresponding kinship pairs across all degrees and unrelated individuals. We systematically evaluated the performance of shared identity-by-descent (IBD) segment analysis, likelihood ratio (LR) method, and kinship coefficient approaches. Innovatively, we developed a multi-feature ML model incorporating IBD distribution statistics, LR values, and kinship coefficients. Concurrently, we created an IBS-CNN model for image recognition by integrating computer vision with identity-by-state (IBS) theory. With pairwise SNP genotypes color-encoded by their IBS status, an IBS-CNN model for IBS heatmap recognition was created. Results demonstrate that the IBS-CNN model ( https://github.com/Rarapie/IBS-CNN ) showed advantages over the common methods and ML models: it directly processes images converted from VCF files while exhibiting greater robustness to data heterogeneity. The IBS-CNN approach achieved robust real-data accuracy (94.56%) while supporting input of IBS barcode heatmaps that exclusively represent IBS status of paired samples, establishing a data transmission format more secure and storage-efficient than raw DNA data.

PMID:
42319452
Bibliographic data and abstract were imported from PubMed on 19 Jun 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.