Non-B DNA structure discovery and function assessment using nanopore translocation times

Abstract The canonical double helical structure of DNA (B-DNA) allows for stability and high-fidelity replication, but emerging evidence suggests an increasingly important role for non-canonical (or non-B) DNA structures in the cell. Non-B DNA structures have been (a) implicated in genetic disorders and carcinogenesis due to their increased mutability, (b) associated with variability in promotor activity, and (c) established as a substrate and site of localization for human telomerase and a putative mechanism for specifying centromere identity. Experimental discovery of non-B DNA structures requires specialized, low-throughput protocols that can be prohibitively expensive and typically target a subset of non-B DNA types. Computational methods offer a lower-cost and higher throughput alternative, but rely on DNA base motifs resulting in two fundamental limitations: (1) few regions having non-B DNA motifs are occupied by non-B DNA structures at any point in time, and thus non-B DNA structure predictions have high false-positive rates that typically limit analyses to aggregate statistics; and (2) non-B DNA structures without well-characterized DNA base motifs are missed entirely. The inability to reliably and efficiently predict non-B DNA structures is a major scientific gap that delays the characterization non-B DNA’s role in medically relevant biological processes and disease. Recently, we discovered that non-B DNA motifs are associated with significant variability in the DNA translocation speeds of nanopore sequencing. We developed predictive models to infer where non- B structure is likely to form among locations occupied by non-B DNA motifs using translocation times as features and performed extensive in silico validation of the predicted structures. While these preliminary analyses demonstrated the feasibility of non-B DNA structure prediction from nanopore sequencing, several significant technical limitations and scientific knowledge gaps limit the tool’s generalizability and performance; we propose to resolve these issues by pursuing two aims: (1) predict and validate non-B DNA structures using nanopore translocation times and other molecular sequencing dynamics; and (2) assess the functional relevance of non-B DNA structures for clinically relevant phenotypes. Successful completion of this project will provide software tools that enable efficient and precise characterization of non-B DNA structures at the single sample level from third-generation sequencing platforms. The proposed research will also generate novel associations between non-B DNA structures and clinically relevant phenotypes. Well- documented and open-source tools will be made easily accessible for usage in new study designs or to reanalyze existing third-generation sequencing data.