摘要
混合DNA图谱解析是法医物证鉴定的核心难题,尤其在面对低模板、降解及多贡献者混合等复杂情形时,现有概率基因分型方法常面临计算效率低与极端不平衡样本解析不准的双重瓶颈.针对上述问题,为精确解析复杂的混合DNA图谱,本文提出一种分阶段推断框架.首先,建立基因座特异性的自适应贝叶斯模型,该模型通过数据校正、混合概率建模,稳健地推断各贡献者的混合比例.其次,利用该比例作为先验信息,构建全局最小残差优化模型,并采用动态规划结合集束搜索的高效算法,从耦合信号中反解出每个贡献者的基因型.该框架将复杂的溯源问题有效解耦,为混合DNA图谱的自动化精准解析提供了新方法.
The interpretation of mixed DNA profiles remains a critical challenge in forensic identification,particularly when confronting complex scenarios characterized by low-template DNA,degradation,and multi-contributor mixtures.Existing Probabilistic Genotyping(PG)methods often suffer from dual bottlenecks:low computational efficiency and insufficient accuracy in resolving extremely unbalanced samples.To address these issues and achieve precise interpretation of complex DNA mixtures,this paper proposes a staged inference framework.First,a locus-specific adaptive Bayesian model is established,which robustly infers the mixture proportions of contributors through data calibration and probabilistic modeling.Second,utilizing these proportions as prior information,a global minimum residual optimization model is constructed.An efficient algorithm combining dynamic programming with beam search is then employed to resolve the genotypes of each contributor from the coupled signals.By effectively decoupling the complex source attribution problem,this framework provides a novel approach for the automated and precise interpretation of mixed DNA profiles.
作者
章一舟
崔紫洋
金湛
陈丕炜
ZHANG Yizhou;CUI Ziyang;JIN Zhan;CHEN Piwei(School of Mathematical Sciences,Ocean University of China,Qingdao,Shandong 266100,China)
出处
《数学建模及其应用》
2025年第4期70-79,共10页
Mathematical Modeling and Its Applications
基金
山东省本科教学改革研究重点项目(Z2024072)
中国海洋大学本科教育教学研究重点项目(2024ZD26,2025D17)。
关键词
混合DNA图谱
比例推断
基因型溯源
贝叶斯推断
动态规划
mixed DNA profile
mixture ratio inference
genotype deconvolution
Bayesian model
dynamic programming
