PSE-HMM: genome-wide CNV detection from NGS data using an HMM with Position-Specific Emission probabilities

BMC Bioinformatics

Methodology Article

Mehdi Sadeghi¹  Seyed Amir Malekpour²  Hamid Pezeshk³ 

[1] National Institute of Genetic Engineering and Biotechnology, Tehran, Iran;School of Mathematics, Statistics and Computer Science, College of Science, University of Tehran, 14155-6455, Tehran, Iran;School of Mathematics, Statistics and Computer Science, College of Science, University of Tehran, 14155-6455, Tehran, Iran;School of Biological Sciences, Institute for Research in Fundamental Sciences (IPM), Tehran, Iran;

关键词: Next Generation Sequencing (NGS);    Hidden Markov Models (HMMs);    Expectation Maximization (EM) algorithm;    mixture densities;    Copy Number Variation (CNV);   

DOI  :  10.1186/s12859-016-1296-y

 received in 2016-05-22, accepted in 2016-10-20,  发布年份 2016

来源: Springer

【 摘 要 】

BackgroundCopy Number Variation (CNV) is envisaged to be a major source of large structural variations in the human genome. In recent years, many studies apply Next Generation Sequencing (NGS) data for the CNV detection. However, still there is a necessity to invent more accurate computational tools.ResultsIn this study, mate pair NGS data are used for the CNV detection in a Hidden Markov Model (HMM). The proposed HMM has position specific emission probabilities, i.e. a Gaussian mixture distribution. Each component in the Gaussian mixture distribution captures a different type of aberration that is observed in the mate pairs, after being mapped to the reference genome. These aberrations may include any increase (decrease) in the insertion size or change in the direction of mate pairs that are mapped to the reference genome. This HMM with Position-Specific Emission probabilities (PSE-HMM) is utilized for the genome-wide detection of deletions and tandem duplications. The performance of PSE-HMM is evaluated on a simulated dataset and also on a real data of a Yoruban HapMap individual, NA18507.ConclusionsPSE-HMM is effective in taking observation dependencies into account and reaches a high accuracy in detecting genome-wide CNVs. MATLAB programs are available at http://bs.ipm.ir/softwares/PSE-HMM/.

【 授权许可】

CC BY   
© The Author(s). 2016

【 图 表 】

12864_2017_3990_Article_IEq12.gif

12864_2017_3777_Article_IEq4.gif

12864_2016_3425_Article_IEq8.gif

12864_2017_3733_Article_IEq43.gif

12864_2017_4020_Article_IEq6.gif

12864_2016_2388_Article_IEq4.gif

12867_2017_86_Article_IEq1.gif

12864_2017_3920_Article_IEq5.gif

12711_2017_331_Article_IEq65.gif

12864_2016_3194_Article_IEq3.gif

12864_2017_3492_Article_IEq4.gif

12864_2017_3605_Article_IEq5.gif

12864_2017_3798_Article_IEq1.gif

12864_2017_3661_Article_IEq5.gif

12864_2017_3605_Article_IEq4.gif

12888_2017_1559_Article_IEq1.gif

12864_2017_4132_Article_IEq31.gif

12888_2017_1528_Article_IEq1.gif