Fig. 2

Illustration of the variant loci vectorization. a Example of variant locus vectorization over 7 target regions of 2 transcripts. 4 mutations are shown as they are mapped on the vectorized coordinates. Two SNVs overlap with both transcripts, the corresponding positions on targets 1,2 for 1st SNV and targets 6,7 for 2nd SNV are set to “1” for the vector that is corresponding to the alternate alleles of the SNVs. Similarly, 2 base-pair deletion is mapped to the vectorized positions impacted by the deletion and 2 consecutive entries in the vector are set to “1”. For the insertion, the position corresponding to the insertion is set “1”. b Example of transcript-specific (Top) and gene-specific (Bottom) target regions. The gene-specific regions do not contain one exon that is redundant in terms of variant annotation. c Illustration of the notations. d Illustration of the multiplication between the impact vector (58-bit annotation value vector) and the variant locus vector. The multiplication is indicated by the cross and the results vector is shown on the right. Any non-zero entry is illustrated by red rectangles. e Example decoding of the junction sequence using the neighborhood sequences for a 3-nucleotide deletion. The junction sequence is formed simply by joining the left and right neighborhood sequences of the first and last nucleotides of the deletion. f The 58-bit packing of variant annotation