基于HSMM和K-means的纳米孔多级事件检测

Detecting of Multi-level Events in Nanopore Experiments Based on HSMM and K-means

提出了一种基于隐半马尔科夫模型(HSMM)和K-means的混合算法,用于对纳米孔数据中出现的多级事件进行分析。通过K-means对检测到的阻断事件进行分析,确定纳米孔信号中的电流幅值水平个数,将其作为HSMM模型的隐状态输入;然后采用HSMM识别多级事件内部的不同水平,确定各水平的幅值和持续时间。最后将算法与Opennanopore软件进行比较,对三级、四级事件进行处理。仿真结果发现:HSMM比Opennanopore软件的准确率高,能够准确对事件进行分级。

This paper proposes a HSMM and K - m e a n s based hybrid algorithm for identifying the multi-level events in nanopore data. B y m e a n s of K - m e a n s to analyze the detected blockade events,the n u m b e r of current amplitude level of nanopore data can be identified,which will be taken as the input of the H SMM’s number of hidden states. And then, HSMM is used to identify different level inside multi-level event and deceide the dwell times and amplitude of different levels. In the situation of third-level and fourth-level event,the simulation results s h o w that HSMM is m o r e accurate than Opennanopore software, and can classify the events accurately.