應用深度神經網路進行基於STR和SNP的DNA混合物解釋

Abstract

短縱列重複序列 (short tandem repeat; STR) 和單一核苷酸多型性 (single nucleotide polymorphism; SNP) 已廣泛用於 DNA 混合物解釋的臨床研究和法醫應用，例如透過 STR 和 SNP 信息識別個體。 在本論文中，我們提出了一個可解釋的 DNA 混合物解釋模型。透過識別重要位點並聯合考慮STR和SNP信息來提高解釋準確性。此外，我們還將這種方法應用於“觸摸樣本”(Touch sample)，指透過觸碰後殘留在物體上極少量的DNA。如結果所示，我們得知此方法是有效的。

Short tandem repeat (STR) and single nucleotide polymorphisms (SNP) have been widely used in clinical research and forensic applications for DNA mixture interpretation, identifying individuals through STR and SNP information. In this thesis, we propose an explainable model for DNA mixture interpretation. Through identifying important locus and jointly considering STR and SNP information to improve the interpretation accuracy. We also applied this method to “touch sample”, which contains extremely small amounts of DNA that left on an object after it has been touched. From the results, we learned that our method is effective.