Unique amino acid sequences at the junctions of fusion or truncated proteins translated from chimeric RNAs form neoantigen peptide sites capable of inducing anti-tumor immune responses. Our objective is to find recurrent fusion transcripts that have the potential to generate candidate neoantigens presented by major histocompatibility complex class I (MHC I) during breast cancer progression. We comprehensively characterized the landscape of fusion transcripts in 225 samples of breast tumors representing 3 subtypes. For each patient we tested 4 sites including Tumor (T), Adjacent Normal (Adj-NL), Distant Normal (Dist-NL-2 sites). Using breast tissue from unaffected individuals (NL) we uncovered 20 novel fusion transcript variants detected from RNAseq data analyzed through two fusion callers. NSF-LRRC37A3 the fusion transcript with the largest number of junction crossing reads per sample and the highest recurrence was selected for further study. The fusion was detected in ~20% of the 75 tumor samples (TNBC=4/25, HER2+=7/25 and HR+=7/25), 5 samples in the TCGA breast cancer dataset and absent in NL (n=4). Some fraction of the patients that presented with NSF-LRRC37A3 in the tumor also contained the fusion in Adj-NL and/or Dist-NL. Interestingly, some patients who were fusion negative for the tumor scored fusion positive for the matched Adj-NL and Distant-NL. The 5'- and 3'-boundaries were found located on the coding strands of Exon 13 of NSF and Exon 5 of LRRC37A3. The two major ORFs were predicted including NSF-Exon 1-13-KFPRKLYFLH (NSF with a C-terminal truncation) and MISNQN-LRRC37A3 Exon 5-15 (LRRC37A3 with an N-terminal truncation). The 2 ORFs were analyzed through MHCnuggets, a deep neural network method that predicts peptide–MHC binding to MHC class I/II. A total of 18 different 8-11 mer neoantigen peptides discovered from the fusion ORFs were predicted to bind to a total of 30 unique MHC class I alleles with a binding affinity of IC50<500nM.