近红外光谱结合支持向量机对STR基因座的化学模式识别分型

Chemical Pattern Recognition Genotyping of STR Locus Based on Nearinfrared Spectroscopy Combined with Support Vector Machine

利用近红外光谱(NIRS)结合支持向量机(SVM)模式识别原理建立了微卫星(短串联重复序列,STR)的分型方法。以D16S539基因座的10-10、10-11和11-11 3种不同基因型为例,研究了包含该多态性位点的脱氧核糖核酸(DNA)片段的聚合酶链反应(PCR)扩增条件和NIRS检测条件,建立了标准化扩增条件和检测条件,以此条件获得了基因分型的标准物及其标准NIRS。以标准光谱为识别变量,建立该3类基因型的SVM判别模型,有效地克服了光谱共线性,并对少样本数表现出良好的稳定性,模型预测率100%。该方法不需任何检测前处理,而只需一步PCR扩增和NIRS检测即可实现STR分型,具有简单、快速、低成本等优点。

By using the near-infrared spectra of short tandem repeats （STR＇s） and applying the principle of support vector machine （SVM） chemical pattern recognition, model for distingushing STRs of different genotypes was established. Taking 3 kinds of genotype, k e. , 10-10, 10-11 and 11-11, of the Locus D16S539 as examples, conditions of NIRS detection and amplification of polyenzyme chain reaction （PCR） of DNA fragment containing polymorphism sites were studied, resulting the establishment of standardized eonditions of amplification and detection, as well as obtaining standard materials and their standard NI1R spectra for genotyping. By taking the NIR spectra as identifying variables, model for distiguishing STRIs of the 3 genotypes mentioned above was established, thus the spectral collinearity was effectively overcome, and good stability was obtained for small number of samples, attaining the model prediction rate of 100%. The present method was proved to have the advantage of achieving the STR genotyping by merely the PCR amplification and NIRS detection, in a simple rapid and economic way, without the necessity of any pretreatment in the detection.