Additional file 1: Figure S1. Workflow to compare bioinformatics tools on Ensembl and RNA-seq data for the predictions of branch point (BP). Figure S2. The different ways that a variant may alter the branch point score. Figure S3. Running time of the four tools SVM-BPfinder, BPP, Branchpointer, and LaBranchor. Figure S4. Paired comparison of the five tools from the Ensembl data and from the RNA-seq data. Figure S5. The overlap of natural 3′ ss (True Calls) and controls AG (False Calls) from Ensembl data. Figure S6. Splicing junctions filtered out from RNA-seq data. Alt 3’ss: alternative acceptor splice sites. Figure S7. The distribution of the relative expression of alternative 3’ss. Figure S8. The overlap of alternative 3′ ss (True Calls) and controls AG (False Calls) from our RNAseq data. Figure S9. Correlation between the scores (SVM-BPfinder, BPP, Branchpointer, LaBranchoR, RNABPS) and the expression of alternative 3’ss. Figure S10. Repartition of variants (n = 120) according their position relative to the predicted branch point. Figure S11:. Determination of optimal motif (YTRAYNN) length to predict splicing alteration, n = 120 variants. ACC: Accuracy, Pos: relative position in branch point motif, Se: Sensitivity, Sp: Specificity. Figure S12. Cross-validation (1000 times) to select the optimal model to predict branch point alteration. Figure S13. Cross-validation (1000 times) to select the optimal model to predict branch point alteration without the positions of predicted BP for all tools except BPP.