ELMSI

ELMSI detects micro-satellite instability (MSI) events from next-generation sequencing data by estimating length distributions and states of micro-satellite regions in mixed tumor samples with paired controls.

Key Features:

• Tumor Purity Estimation: Estimates tumor purity using read counts from filtered single nucleotide variant (SNV) loci to inform downstream analyses.
• Short Micro-satellite Length and State Identification (MLE): Identifies length distributions and states of short micro-satellites (shorter than one read length) via Maximum Likelihood Estimation (MLE).
• Long Micro-satellite Inference (EM + CLT): Infers length distributions of micro-satellites up to 10 kilobases using a simplified Expectation Maximization (EM) algorithm combined with the central limit theorem.
• Statistical State Determination: Applies statistical tests to estimated length distributions to determine micro-satellite states.
• Benchmarking: Validated against MSIsensor for shorter micro-satellites and assessed for performance on longer regions across mixed-purity samples.

Scientific Applications:

• MSI detection in cancer genomics: Detects and characterizes micro-satellite instability in tumor samples using next-generation sequencing.
• Mixed-sample and variable-purity analysis: Enables accurate estimation of micro-satellite lengths and states in mixed tumor/control samples with varying tumor purity.
• Support for downstream tumor genetics analyses: Provides quantitative microsatellite length and state estimates to inform tumor genetics studies and clinical decision-making.

Methodology:

Estimating tumor purity from read counts at filtered SNV loci; Maximum Likelihood Estimation (MLE) for short micro-satellites; simplified Expectation Maximization (EM) with the central limit theorem for long micro-satellites up to 10 kilobases; statistical tests for state determination.