Variant calling

Variant calling is the process of identifying positions in a DNA sample where the sequenced data differs from a reference genome, single nucleotide changes, insertions, deletions, or larger structural differences, collectively called variants.

It starts from reads already mapped to a reference genome through sequence alignment, then applies statistical models to distinguish genuine variants from sequencing errors, accounting for read depth, base quality, and, for diploid organisms, the expectation that most positions will show either a heterozygous or homozygous signal rather than something in between. The output is typically a standardized list of variants, each carrying a confidence score.

Variant calling underlies clinical genetic testing, cancer genomics, and population studies, and when variants are compared across many samples it feeds directly into phylogenetic analysis to trace relationships between individuals or lineages. Pitfalls include low sequencing depth or poor alignment quality in repetitive regions, which produces both false positives and missed real variants, and the gap between a statistically confident call and a clinically meaningful one, interpreting what a variant actually does is a separate, further step beyond calling it.