A multiple predictive tool approach for phenotypic and biogeographical ancestry inferences

Phenotypic and biogeographical ancestry predictions for 128 Canadians with various self-reported ancestries, generated using the ForenSeq™ DNA Signature Prep kit/Primer Mix B and VEROGEN’s Universal Analysis Software (UAS) v1.3, were compared to predictions derived from the ERASMUS v2.0, Snipper Application v2.5, Forensic Reference/Resource on Genetics knowledge base [FROG-kb] ©2019 and STRUCTURE v2.3.4, web tools. Performance metrics were calculated for all tools tested. The UAS v1.3 eye color predictions were determined to be accurate (91.8% for brown, 82.4% for blue) with no need to complement results with any other predictive tools. For hair color predictions, the UAS v1.3 was accurate for black (93.5%), for red (80.0%) but limited for brown (50.0%) and for blond (31.3%) using the Highest Probability Approach. The ERASMUS web tool could complement the UAS results using a revised Prediction Guide Approach. Non-admixed individuals with ancestries from Africa, East Asia or Europe were predicted with 79.7% accuracy using the UAS v1.3. Admixed individuals and those with ancestries from India, the Middle East and South America were better classified using Snipper, FROG-kb and STRUCTURE. Complementing UAS v1.3 predictions with those obtained from open access web tools thus represents a way to maximize information derivable from unknown samples.本研究针对128名具有不同自我报告血统的加拿大人群，使用ForenSeq™ DNA Signature Prep试剂盒/引物混合液B与VEROGEN通用分析软件（Universal Analysis Software, UAS）v1.3生成表型与生物地理祖先血统预测结果，并将其与ERASMUS v2.0、Snipper Application v2.5、法医学遗传学参考资源知识库（Forensic Reference/Resource on Genetics knowledge base, FROG-kb）©2019以及STRUCTURE v2.3.4等网络工具得到的预测结果进行对比。本研究对所有测试工具均计算了性能指标。经评估，UAS v1.3的虹膜颜色预测准确率可达：棕色眼睛91.8%、蓝色眼睛82.4%，无需借助其他预测工具补充结果。在头发颜色预测方面，采用最高概率法（Highest Probability Approach）时，UAS v1.3对黑色头发（93.5%）、红色头发（80.0%）的预测准确率较高，但对棕色头发（50.0%）与金色头发（31.3%）的预测效果有限。ERASMUS网络工具可通过改进后的预测指南法（Prediction Guide Approach）对UAS的预测结果进行补充。使用UAS v1.3对来自非洲、东亚或欧洲的非混血个体进行血统预测时，准确率可达79.7%。而对于混血个体以及来自印度、中东与南美洲的血统个体，采用Snipper、FROG-kb与STRUCTURE工具进行分类的效果更佳。因此，借助开源网络工具的预测结果对UAS v1.3的预测结果进行补充，是最大化从未知样本中获取可用信息的有效途径。