the success of checkpoint inhibitors in cancer therapy is largely attributed to activating the patient's immune response to their tumor's neoantigens arising from DnA mutations. this realization has motivated the interest in personal cancer vaccines based on sequencing the patient's tumor DnA to discover neoantigens. Here we propose an additional, unrecognized source of tumor neoantigens. We show that errors in transcription of microsatellites (MS) and mis-splicing of exons create highly immunogenic frameshift (fS) neoantigens in tumors. the sequence of these fS neoantigens are predictable, allowing creation of a peptide array representing all possible neoantigen fS peptides. this array can be used to detect the antibody response in a patient to the FS peptides. A survey of 5 types of cancers reveals peptides that are personally reactive for each patient. this source of neoantigens and the method to discover them may be useful in developing cancer vaccines. Checkpoint inhibitor immunotherapeutics are revolutionizing cancer therapy. However, even in the most responsive cancers a substantial portion (50-80%) of the patients have poor to no positive response 1-5. A surprising finding in the analysis of these patients was that one of the best correlates of response has been the total number of neoantigens in the tumor 6-8. This is also the case for patients with high microsatellite instability (MSI) where the production of FS neoantigens drives the effective anti-tumor immune responses 9-11. The realization of the immunological importance of these DNA mutations has spawned the effort to develop personal vaccines 12. As promising as early studies are of these vaccines, a major problem is that the majority of tumors will not have enough neoantigen-generating mutations to sustain development of a personal vaccine 13-15. For example, melanoma tumors have a high mutational level with an average of 200 neoepitope mutations. This provides a large number to algorithmically screen for optimal antigenic presentation. In recent reports of two Phase I clinical trials of personal melanoma vaccines, starting with 90~2,000 personal neoantigens, 10 or 20 were identified for the vaccine 16,17. However, in glioblastoma multiforme (GBM) only 3.5% patients had a high tumor mutation load, and further analysis showed that only a very small subset of GBM patients would potentially benefit from checkpoint blockade treatment 18. This is also consistent with a lack of response in GBM patients to checkpoint inhibitors 19. Massive genomic sequencing results indicated that GBM, ovarian cancer, breast adenocarcinoma and many other cancer types had very low number non-synonymous mutations, which will make these cancers difficult targets for personalized cancer vaccines 14,20. To solve this problem, we have investigated an alternative source of neoantigens which could possibly expand the scope of the application and efficacy of the neoantigen based cancer vaccines. In the process of becoming a tumor, not only does the DNA mutation rate increase wit...