A Hidden Markov Model for Copy Number Variant prediction from whole genome resequencing data

Abstract

MotivationCopy Number Variants (CNVs) are important genetic factors for studying human diseases. While high-throughput whole genome re-sequencing provides multiple lines of evidence for detecting CNVs, computational algorithms need to be tailored for different type or size of CNVs under different experimental designs.ResultsTo achieve optimal power and resolution of detecting CNVs at low depth of coverage, we implemented a Hidden Markov Model that integrates both depth of coverage and mate-pair relationship. The novelty of our algorithm is that we infer the likelihood of carrying a deletion jointly from multiple mate pairs in a region without the requirement of a single mate pairs being obvious outliers. By integrating all useful information in a comprehensive model, our method is able to detect medium-size deletions (200-2000bp) at low depth (<10× per sample). We applied the method to simulated data and demonstrate the power of detecting medium-size deletions is close to theoretical values.AvailabilityA program implemented in Java, Zinfandel, is available at http://www.cs.columbia.edu/~itsik/zinfandel/

Detection of Microdeletion 22q11.2 in a Fetus by Next-Generation Sequencing of Maternal Plasma

“…To detect this 3-Mb fetal deletion in our study, we used a genomic coverage of 4-fold-an increase in coverage of approximately 20-fold over that of current standard aneuploidy detection. Smaller deletions (potentially down to 0.5 Mb) or samples containing less fetal ccf DNA will require even higher coverage (15 ). Consequently, massively parallel shotgun sequencing will have to provide substantially higher throughput, or sequencing will have to be performed with a cost-effective targeted sequencing methodology before this method can be considered for routine clinical use.…”

Simple and efficient germline copy number variant visualization method for the Ion AmpliSeq™ custom panel

“…In addition, we employed the t ‐test to identify statistically different normalized relative read depths for five continuous amplicons compared to the normalized relative read depth of all amplicons to aid the manual inspection. For further statistical copy number variation detection, please refer to other methods using the Hidden Markov Model to estimate the copy number from observed data if necessary (Shen, Gu, & Pe'er, ).…”

Changepoint Analysis for Efficient Variant Calling

“…A similar idea was used by Shen and Zhang [9] to detect abrupt changes in copy number. One way to view these methods is as an alternative to the hidden Markov model (HMM), which has also been previously used to detect genomic variation [10,11]. In contrast to the latter methods, which require the number of hidden states to be known a priori, changepoint methods allow the number of detected segments to vary in accordance with the data.…”

“…We used the same window size for AllChange. For CAGe++, we set the variant calling parameters to (α 1 , α 2 , α 3 , α 4 , α 5 ) = (12,10,20,3,20).…”