RT Journal
A1 Shen, Yufeng
A1 Wan, Zhengzheng
A1 Coarfa, Cristian
A1 Drabek, Rafal
A1 Chen, Lei
A1 Ostrowski, Elizabeth A.
A1 Liu, Yue
A1 Weinstock, George M.
A1 Wheeler, David A.
A1 Gibbs, Richard A.
A1 Yu, Fuli
T1 A SNP discovery method to assess variant allele probability from next-generation resequencing data
JF Genome Research 
JO Genome Research 
YR 2009 
FD December 17 
DO 10.1101/gr.096388.109 
UL http://genome.cshlp.org/content/early/2009/12/16/gr.096388.109.abstract 
AB Accurate identification of genetic variants from next-generation sequencing (NGS) data is essential for immediate large-scale genomic endeavors such as the 1000 Genomes Project, and is crucial for further genetic analysis based on the discoveries. The key challenge in single nucleotide polymorphism (SNP) discovery is to distinguish true individual variants (occurring at a low frequency) from sequencing errors (often occurring at frequencies orders of magnitude higher). Therefore, knowledge of the error probabilities of base calls is essential. We have developed Atlas-SNP2, a computational tool that detects and accounts for systematic sequencing errors caused by context-related variables in a logistic regression model learned from training data sets. Subsequently, it estimates the posterior error probability for each substitution through a Bayesian formula that integrates prior knowledge of the overall sequencing error probability and the estimated SNP rate with the results from the logistic regression model for the given substitutions. The estimated posterior SNP probability can be used to distinguish true SNPs from sequencing errors. Validation results show that Atlas-SNP2 achieves a false-positive rate of lower than 10%, with an ∼5% or lower false-negative rate.