基于SNP数据检测染色体拷贝数结果可信度分析

Confidence level analysis of detecting chromosome copy number detection using SNP array

利用SNP数据检测肿瘤细胞染色体拷贝数变异是癌症相关研究的一个热点,目前已有多种方法可以通过分析SNP array数据检测染色体拷贝数。然而在某些情况下,这些检测方法检测结果与真实拷贝数具有一定错误率。目前并没有方法研究预测结果发生错误的规律。本文分别分析了GPHMM,ASCAT两种检测方法结果信息熵与检测正确率的关系,发现检测正确率与信息熵存在很强的相关性。通过对比不同肿瘤细胞比例下信息熵与正确率关系,本文发现随着肿瘤细胞比例的增大,检测结果信息熵平均值增大,方差减小;同时平均检测正确率也越来越大,方差显著减小。这些结果显示信息熵的大小可以反映出检测结果正确率的高低。最后,本文以高肿瘤细胞比例下拷贝数检测结果为例,研究了在变异类型单一,信息熵小的情况下,染色体倍性检测的正确率。结果表明信息熵可以作为衡量检测结果可信度的指标:即信息熵越高,检测结果越可信。

Recently,using SNP arrays to detect chromosomal copy number aberrations of tumor cells gains its popularity.Several methods that devoted for copy number dissection have been proposed. However,there is no study being performed regarding the error rate of results of copy number detection comparing with true copy number profile. In this study,by using GPHMM and ASCAT,which are both devoted for copy number detection,examinations on the relationship between entropy and accuracy are conducted and results show that accuracy and entropy demonstrate a strong correlation. By testing the accuracy and entropy under different tumor cell proportions,results show that with the increase of the proportion of the tumor cells,average entropy of detection results become larger and the variance becomes smaller. Also,study finds that the average rate of correct detection is significantly increasing when the variance is decreasing,indicating that the proportion of tumor cells can affect the accuracy of detection and information entropy at the same time. At last,by taking an error detection case of tumor samples with high proportion of tumor cells,study shows that limited kinds of aberrations and small entropy are likely to cause the occurrence of serious bias in average copy number estimation. In conclusion,all results suggest that entropy can act as confidence level indicator for copy number detection： the higher entropy is likely to produce the better reliability regarding copy number detection.