Long neoantigens generated by frameshift mutation are associated with good prognosis in patients with cancer. (A) Genomic organization of the WT and targeting vectors for the murine Magd1 gene. Bpmel-Magd1 WT knock-in was generated by inserting GFP, while Bpmel-Magd1 FS1 knock-in was generated by inserting GFP along with 2 adenines to induce a downstream type 1 frameshift (FS1). Similarly, Bpmel-Magd1 FS2 knock-in was generated by inserting GFP plus 1 adenine to induce a downstream type 2 frameshift (FS2). (B) The experimental procedure for anti–PD-L1 therapy in cell mixture tumors (left panel). The FS1 mixture was created by combining 50% Bpmel-Magd1 FS1 knock-in cells with 50% Bpmel-Magd1 WT knock-in cells, while the FS2 mixture was generated by mixing 50% Bpmel-Magd1 FS2 knock-in cells with 50% Bpmel-Magd1 WT knock-in cells. The right panel shows the tumor volume of the cell mixture following anti–PD-L1 therapy (n = 5 per group). P value was determined using one-way ANOVA followed by Dunnett’s multiple comparisons test. (C) Waterfall plot for the top 20 mutated genes in 436 colorectal cancer patients based on the TCGA dataset. (D) Percentages of different lengths of frameshift-generated functional peptides in the patients. (E) Kaplan‒Meier analysis of the overall survival of patients whose maximal length of frameshift-generated functional peptides was longer (long FP) or shorter (short FP) than 120 AA (left). FS, frameshift. The top 20% were defined as having high TMB or FS numbers, corresponding to the frequency of patients with long functional peptides. The Kaplan‒Meier method was used to compare two groups of clinical variables. Experiments in B were performed three times.