Abstract Background: Triple-negative breast cancer (TNBC) is a heterogeneous disease. Clinically, we observe three distinct TNBC outcomes: 1) rapid relapse (rrTNBC) characterized by aggressive drug resistant disease; 2) late relapse (lrTNBC) characterized by indolent or treatment responsive disease; and 3) no relapse (NoRTNBC). We hypothesized that distinct clinical and genomic features of primary tumors define rapid versus late relapse in TNBC. Approach: Using three publicly-available datasets (METABRIC, TCGA, and a prior gene expression meta-analysis), we identified 455 patients diagnosed with primary TNBC with adequate follow-up to be characterized as rrTNBC (relapse or death within 2 years of diagnosis), lrTNBC (relapse or death more than 2 years after diagnosis), or NoRTNBC (no relapse/death with at least 5 years follow-up). We compiled basic clinical (n=455 patients) and primary tumor multi-omic data, including whole transcriptome (n=455), whole genome copy number (n=317), and mutation data for 171 cancer-related genes (n=317). We evaluated intrinsic subtypes (PAM50, TNBCtype), 125 gene expression signatures, CIBERSORT immune subsets, copy number, and mutation frequency. Results: We first evaluated patients with relapse (rrTNBC+lrTNBC) vs. NoRTNBC. There was no significant difference in age, grade, stage at diagnosis, or PAM50 or TNBC subtype proportion between relapse and NoRTNBC. Among 125 expression signatures, five immune signatures were significantly higher in NoRTNBCs (FDR p<0.05) suggesting increased immune activity in patients who do not relapse. Using CIBERSORT inferred immune subsets, anti-tumor CD8 T-cell, M1 macrophage, and gamma-delta T-cell subsets were all highly correlated to these immune signatures (all Pearson's r >= 0.3, all p<1.2e-8). Among genomic features, patients who relapsed were significantly more likely to harbor a mutation in PIK3CA (Fisher exact FDR p=0.02) but there was no significant difference in tumor mutation burden or percent genome altered (Student's t-test p=0.83 and p=0.99, respectively). We then evaluated primary TNBC genomic data in patients who ultimately developed rapid vs. late relapse. Patients with rrTNBC were more likely to be higher stage (p<0.0001) while lrTNBC were more likely to be non-basal PAM50 subtype (p=0.03). Among 11 significantly altered gene expression signatures (FDR p<0.05), 6 estrogen/luminal signatures were significantly higher in lrTNBC. Mutations in DNAH11 and PIK3CA were more common in lrTNBC (Fisher exact FDR p=0.04 and p=0.05, respectively) but there were no significant differences in tumor mutation burden or copy number burden (Student's t-test p=0.13 and p=0.45, respectively). Using 317 cases with full genomic data divided into training and validation datasets, we will report a comparison of machine learning models for predicting relapse versus no relapse and rapid versus late relapse. Conclusions: Primary TNBC tumors destined for rapid, late, or no relapse reflect distinct genomic features. Anti-tumor immune signatures and subsets are enriched in patients who do not relapse yet no difference in mutational or copy number burden. Relative to rapid relapse TNBCs, late relapse TNBCs are enriched for non-basal tumors, estrogen/luminal expression signatures, and mutations in DNAH11 and PIK3CA. Citation Format: Zhang Y, Nock W, Asad S, Adams E, Singh J, Damicis A, Lustberg MB, Noonan A, Reinbolt R, Sardesai S, VanDeusen J, Wesolowski R, Williams N, Ramaswamy B, Stover DG. Multi-omic predictor of rapid and late relapse in primary triple negative breast cancer [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P3-07-08.