A kinetic model-based algorithm to classify NGS short reads by their allele origin.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next Generation Sequencing (NGS) enables large-scale genetic variant discovery.
• Accurate zygosity assignment is crucial for variant interpretation in diploid genomes.
• Existing genotyping methods often rely on probabilistic models and a priori knowledge.

Purpose of the Study:

• To introduce Kimimila, a novel algorithm for accurate variant genotyping from NGS data.
• To evaluate Kimimila's performance against established genotyping tools.
• To demonstrate Kimimila's ability to assign sequencing reads to specific alleles.

Main Methods:

• Developed Kimimila (KInetic Modeling based on InforMation theory to Infer Labels of Alleles), a distance geometry-based algorithm.
• Assessed performance using simulated and real data from the 1000 Genomes Project.
• Compared Kimimila against GATK, Samtools, VarScan, FreeBayes, and Atlas2.

Main Results:

• Kimimila achieved comparable accuracy in correctly genotyped variants to established methods.
• The algorithm accurately assigns sequencing reads to specific alleles without prior assumptions.
• Kimimila facilitates the splitting of read pools based on inferred allele origin.

Conclusions:

• Kimimila offers a robust, assumption-free approach to variant genotyping in NGS data.
• The method provides accurate genotype inference and read assignment capabilities.
• Kimimila enhances the analysis of genetic variation in diploid genomes.