Fig 1: SPICE enables the design of cancer mutation-specific splicing elements.a, Schematic of the integrated experimental–computational framework for designing splicing sequences specific to cells harboring splicing factor mutations. Cell lines carrying defined splicing factor mutations were first generated, and a pooled minigene library was introduced into paired wild-type (WT) and mutant cell lines. RNA was harvested, and isoform proportions were quantified by sequencing. These data were used to train the Melange model, which subsequently designed new cell type-specific sequences. The designed sequences were synthesized, reintroduced into the same WT and mutant cell lines, and experimentally validated to identify mutation-specific splicing elements.b, Splicing factor mutations profiled in this study, their associated cancer types, and characteristic splicing alterations.c, Volcano plots of significantly differentially spliced library elements (ΔPSI) for each mutant versus wild-type cell line. Significant events have ΔPSI > 0.1 and FDR < 0.05. Red points indicate higher PSI in knockout (KO) cells than in WT cells, and blue points indicate lower PSI in KO cells.d, Number of significant library elements identified per mutant background, filtered by ΔPSI > 0.1 and FDR < 0.05.e, Heatmap of PSI values for significant designed sequences specific to RBM5 and RBM10 KO cells (ΔPSI > 0.1, FDR < 0.05).f, Heatmap showing PSI values for representative parent-design pairs across WT and mutant cell lines.g, Representative examples of paired parent and designed sequences. Nucleotides altered relative to the parent sequence are highlighted.