机器学习技术结合Y染色体CpG位点推断男性年龄

Inferring male individual age based on machine learning technology and Y chromosome CpG locus

目的 年龄推断作为个体特征描绘关键环节之一,在法医实践中的地位愈发重要,然而从混合斑中推断出个体年龄是一个尚未解决的难题。本研究通过筛选Y染色体上CpG位点并结合机器学习算法构建男性个体年龄推断模型。方法 从GEO数据库筛选男性血液甲基化数据,按照年龄变化趋势对Y染色体上的位点进行差异性分析。对Y染色体上的差异性CpG位点进行线性回归拟合并计算Spearman相关系数,基于年龄相关CpG位点构建支持向量机、梯度提升机、随机森林、多元线性回归、K最邻近等5种机器学习模型用于年龄推断模拟和验证。结果差异性分析得到26个CpG位点,通过线性拟合得到8个年龄相关CpG位点。五种机器学习训练模型推断年龄的平均绝对偏差(Mean Absolute Deviation,MAD)范围从5.19岁~7.68岁。在验证阶段的数据集中,支持向量机的年龄推断模型表现最佳,MAD为5.43岁。结论 本研究基于机器学习算法验证了Y染色体上的CpG位点对男性个体年龄推断的可行性,为法医实践提供有效依据。

Objective Age inference is becoming increasingly important in forensic practice as one of the key issues of individual characterization,but it is still an unsolved problem to infer individual age from mixed stains.In this study,a male age inference model was established by screening CpG sites on Y chromosome and combining machine learning algorithm.Methods The male blood methylation data sets were screened from the GEO database,and the differences of the sites on the Y chromosome were analyzed according to the age trend.Linear regression was fitted on the different Y-CpGs to calculate the Spearman correlation coefficient.Based on age correlated CpG loci,five machine learning models,including support vector machine,gradient boosting machine,random forest,multiple linear regression,and K-nearest neighbor,were constructed for age inference simulation and verification.Results Analysis of variance yielded 26 Y-CpG,and 8 age related Y-CpG were obtained by linear fitting.The inferred age of the five machine learning training models had the mean absolute deviation(MAD)ranging from 5.19 to 7.68 years.At the validation stage,the support vector machine age inference model performed the best,with a MAD of 5.43 years.Conclusion Based on the machine learning algorithm,this study verified the feasibility of the Y-CpG in inferring the age of male individuals,providing an effective basis for forensic practice.