基于混合泊松分布的新生突变识别算法

对个体而言，不经父母遗传而后天获得的突变称为新生突变，绝大多数癌症都起自新生突变。构建快速精确的变异识别算法将有助于对癌症的研究。然而，针对前期新生突变识别算法准确率不高，且耗时多等问题，本文引入了基于变异位点的先验概率分布模型，运用基于混合泊松分布的期望最大化（EM）算法对新生突变识别算法进行改进与优化，研究了有亲缘关系的新生突变的识别，并在识别精度与运算速度方面与已有算法进行对比。结果表明，基于混合泊松分布的 期望最大化算法在提高运算速度的同时降低了假阳性比率，具有良好的识别效果。

Recognition of de Novo Mutations Based on Hybrid Poisson Distribution

For the individual, gene mutations that are acquired without parental inheritance are the origins of vast majority of cancers. Application of fast and accurate recognition algorithms will be a great help to the study of cancer. Aiming at the problem of poor accuracy and time consumption, a prior probability model of mutation sites was introduced. To modify and optimize the recognition algorithm, the Expectation Maximum (EM) algorithm based on mixed Poisson distribution was used to identify the de novo mutation involving kinship data and compare with the existing algorithms in recognition accuracy and computing speed. The results show that the EM algorithm based on mixed Poisson distribution can improve the speed of operation and reduce the false positive ratio, which is of great significance for the recognition of cancer.