GenoTan

GenoTan identifies length variations in microsatellite alleles from short sequence reads to distinguish true allelic length variants from sequencing- and PCR-induced indel noise.

Key Features:

• Microsatellite genotyping: Accurately identifies length variations in microsatellite alleles using short sequence reads.
• Discretized Gaussian mixture model: Models allele length distributions with a discretized Gaussian mixture model to support precise genotype calls.
• Rules-based approach: Integrates a rules-based component with the mixture model to separate true length variants from noise.
• Bidirectional indel error modeling: Explicitly accounts for the bidirectional nature of insertion/deletion (indel) errors to improve genotyping accuracy.
• Homopolymer decomposition: Implements a homopolymer decomposition method that estimates bias toward insertion or deletion within homopolymer runs.
• Noise source modeling: Models common error sources including PCR amplification errors, individual cell mutations, and misalignment/mis-mapping.
• Performance: Validated on simulated data at 40× coverage with 94.9% genotyping accuracy and reported to be 5–30× faster than competing programs on the same dataset.
• Real-data validation: Evaluated on mixed sequence data from two inbred Drosophila lines and produced the highest true-positive rate among evaluated programs.

Scientific Applications:

• Microsatellite genotyping from short-read sequencing: Enables precise allele length calls for microsatellite loci using short sequencing reads.
• Genotyping in repetitive and homopolymer regions: Targets loci prone to homopolymer-associated indel errors and repetitive-sequence challenges.
• Discrimination of true variants from technical noise: Distinguishes genuine length polymorphisms from PCR- and sequencing-induced indels.
• Analysis of mixed or inbred samples: Applicable to mixed-sample sequencing and inbred-line comparisons, exemplified by Drosophila inbred-line data.

Methodology:

Integrates a discretized Gaussian mixture model with a rules-based approach, incorporates a homopolymer decomposition method to estimate insertion/deletion bias, and explicitly models bidirectional indel errors; validated using simulated 40× coverage datasets and mixed Drosophila inbred-line sequence data.