Abstract Interpreting the role of intergenic sequences in tumor development remains an active area of research critical to understand the overall influence of germline and somatic mutations on disease progression. Recent studies have shown transcription and translation at ostensibly non-coding regions in different cell types, and our own analysis of published datasets across cancer cell lines and patient tumors have revealed extensive intergenic transcription and translation, particularly at or near structural variants (SVs). Given that SVs often fuse genic and intergenic regions together, they are a plausible cause for several intergenic transcription and translation events, and identifying such links would help establish a significant downstream consequence of genomic instability and indicate how intergenic regions could shape tumor biology. In Aim 1, we will develop a machine learning method to estimate the causal effect of SV presence within a locus upon local intergenic transcription and compute such estimates across large public databases of tumor genomes and transcriptomes. This aim will seek to address two outstanding problems in linking somatic mutations to gene expression: (a) non-linear relationships between mutational load and expression magnitude, and (b) the relative rarity of SVs compared to smaller somatic variation that can complicate causal inference. In Aim 2, we will explore the role of nonsense mediated decay (NMD) in shaping the observed expression levels of intergenic and fusion transcripts, particularly if the fusion transcript involves at least one intergenic partner. Building on past studies studying how mutant transcripts evade NMD, we will determine whether such predictions can explain observed variation in expression of SV-linked or standalone intergenic transcripts across samples. By exploring the role of NMD, we will provide new insights into how intergenic and SV-generated transcripts persist within the tumor transcriptome. Finally, in Aim 3, we will analyze the landscape of SV-linked and standalone intergenic transcripts from published datasets in which patients are treated with PD-1 or CTLA-4 checkpoint inhibitors. We will determine whether an increased burden of intergenic transcription is associated with elevated immune infiltration and inflammatory pathway activation and whether SVs that generate intergenic transcripts are linked to improved clinical benefit. We will also test whether expression levels of transcripts predicted to evade NMD are also associated with increased immune responses upon checkpoint inhibition. By analyzing immunotherapy data, we can establish whether SV-linked or standalone intergenic transcription serves as a useful correlate for improved clinical responses. In summary, this study will develop computational approaches to test hypotheses on how intergenic transcripts are generated and maintained and whether they are useful for immunotherapy. This research will contribute new insights into the causes and consequences of intergenic alterations in the tumor genome.