Research Projects

Tumors alter their genomic packaging, chromatin, for their proliferative advantages. Numerous factors from small to large scale contribute to this alteration. A few examples at small scale are CpG methylation, nucleosome positioning, Transcription Factor (TF) binding patterns, post-translational modification of histones, and at large scale, formation of new topologically associated domains and more. Learning these deviations from normal could serve mechanistic insights and help find potential biomarkers. Sequencing DNA molecules captured after targeted or untargeted assays on fragmented chromatin is a common denominator for mapping these events in cells or tissues. But DNA insert size of less than 1000 base pairs and failure to capture epigenetic modification directly are two major limitations of the popular Illumina sequencing technology. By contrast, long-read sequencing technology like Oxford Nanopore addresses both issues and enables mapping protein-DNA binding events at the single-molecule level. With this rationale, I propose to make a significant contribution in deconvolving i) TF-TF interaction mediated cooperativity i.e., indirect TF binding, and ii) cell-type heterogeneity. Indirect binding refers to the TF of interest forming a complex with another TF already bound or binding to DNA. Classic examples are Estrogen Receptor binding to DNA-tethered AP-1 proteins (c-Jun and Fos) in ER+ breast cancers and myocardin binding to SRF. Despite being biologically relevant It is a hard problem to resolve because of the lack of the target’s cognate DNA binding motifs. The idea is to use chromatin dynamics i.e., protein-DNA binding states to significantly refine sequence search space. Briefly, inferring footprints on individual sequenced DNA molecule maps precise and absolute binding status of proteins i.e., TF-bound, nucleosome, and protein-free states on chromatin in vivo. Access to just nucleosome occupancies has helped in previous studies. So, combining existing ChIP-seq with all three states can significantly refine sequence search (Objectives 1 and 2; Preliminary Data). Single-cell RNA-seq is widely used to infer cell types in tumor biopsies. But it is expensive and laborious. But even in bulk assay if we can map stable and reliable DNA methylation patterns on a single-molecule level, we can infer cell types. With long-reads in the nanopore, we can easily span CpG islands and transcription start sites in tandem, a promising combination to accurately infer gene expression status. Similarly, other regions of interest can be used as features. We plan to execute this idea on a pilot system where i) we know distinct methylation patterns at a known set of loci, and ii) we know the percentage contribution of cell types. We will explore Nanopore readouts and device rules for making accurate inferences. This has a promising application in inferring cancer stem-cell populations in biopsies.