SeqEM

SeqEM estimates individual genotypes from next-generation sequencing (NGS) short-read data by adaptively estimating parameters underlying posterior genotype probabilities using an Expectation-Maximization (EM) framework.

Key Features:

• Adaptive Parameter Estimation: SeqEM adaptively estimates parameters underlying posterior genotype probabilities rather than using fixed a priori settings.
• Expectation-Maximization (EM): SeqEM applies the EM algorithm to derive likelihoods from NGS data and to iteratively estimate genotype probabilities and nucleotide-read error rates.
• Use of Sample Information: SeqEM leverages information across a sample of unrelated individuals to improve estimation of genotype probabilities and error rates.
• Empirical Evaluation and Validation: SeqEM's error rates were assessed by analytic calculations and simulations and compared against MAQ and SOAPsnp, and it was applied to exome sequence data from eight related individuals with comparison to genotypes from an Illumina SNP array.

Scientific Applications:

• NGS and exome genotype calling: Accurate determination of genotypes from short-read NGS and exome sequencing data.
• Population genetics and association studies: Precise genotype calls for analyses in population genetics and genetic association studies.
• Validation and benchmarking: Comparison and validation of NGS-derived genotypes against SNP-array genotypes such as those from an Illumina SNP array.
• Personalized medicine: Generation of reliable genotype data to support clinically oriented analyses that require accurate individual genotypes.

Methodology:

SeqEM uses an Expectation-Maximization algorithm within a likelihood framework to iteratively refine estimates of posterior genotype probabilities and nucleotide-read error rates, leveraging information across a sample of unrelated individuals; analytic calculations and simulations were used to evaluate performance.