Fig. 3

Example VCF and jVCF calls for a graph containing nested variation. Three sites exist in the graph, and site-opening edges are labelled by haplogroup (0 or 1). Below the graph, parts of VCF and jVCF files describe a genotyped sample (blue path in the graph). In VCF, only one variant site is genotyped, with the “sample” column stating that allele 4 has been called (this corresponds to “ATCCAC”, as the numbering is 0-based). However, the two nested variants—sites 1 and 2—cannot be expressed independently from site 0, as they do not occur on the linear reference (red path in the graph). jVCF stores the same information as VCF in the “Sites” array (one entry per site) and additionally records site relationships using a “Child_Map” entry, which states that sites 1 and 2 occur under site 0 (first key), haplogroup 1 (second key). This places sites 1 and 2 on an alternate reference background (the sequence “ATACTTC”) allowing them to be expressed in the “Sites” entry. Alternate reference sequences are obtained by following haplogroup 0 at each nested variant site (here this spells “ATACTTC”)