Figure 3.

Calculation of features for the splice acceptor site

(A) SQUIRLS calculates $\Delta R_{ican}$ as the difference in the R_i between the reference and alternate sequence of the canonical acceptor sequence (Figure 1A). If the variant is located outside of this sequence, $\Delta R_{ican}$= 0. SQUIRLS evaluates the potential of variants to create cryptic splice sites using a sliding window approach (material and methods). $\Delta R_{icrypt}$ is calculated by subtracting R_i of the alternate sequence of the canonical acceptor site from the R_i of the best candidate cryptic splice site. The random forest for acceptor variants does not use $\Delta R_{inext}$ as our initial analysis showed that it did not boost classification performance. See Table 1 for information about other features.

(B) In this example, a coding variant activates a cryptic acceptor site with 15 bits which is greater than the canonical acceptor site by 7 bits ($\Delta R_{icrypt}$).

(C) A situation where a variant located in a splice acceptor site introduces novel AG dinucleotide into the AGEZ leading to cryptic splicing or exon skipping.