Unlocking the Hidden Regulome underlying Ocular Health and Disease via Scalable Integrative Frameworks - Project Summary Transcriptional regulators (TRs) play a pivotal role in eye development and disease. Understanding the relationship between TRs and visual function is fundamental in vision research. Advances in next-generation sequencing (NGS) have provided a wealth of omics data from various eye tissues under different conditions, allowing us to pinpoint the tissue-specific TRs underlying both normal vision and eye disease. However, current in-silico TR identification faces significant limitations such as reliance on gene lists or motif information, inadequate handling of within-TR heterogeneity across tissue types and conditions, no uncertainty quantification, outdated reference libraries, and poor scalability. To address these challenges, we propose TRex, a variational Bayesian framework built upon a comprehensive, updated reference library. TRex leverages abundant omics data from various eye tissues to systematically identify TRs associated with vision and eye disease. Understanding the precise function of a TR also requires knowledge of its cell-type-specific binding profiles due to the cellular heterogeneity of ocular tissues. The complex protocols and prohibitive costs of single-cell TR ChIP- seq experiments necessitate the development of novel methods to infer these profiles from readily available bulk TR ChIP-seq data in eye research. We introduce sideTR, an innovative Bayesian multi-instance learning model, which leverages information from single-cell ATAC-seq and matched RNA-seq (when available) to achieve deconvolution. As the first method capable of dissecting bulk TR ChIP-seq data, sideTR will reveal cell-type- specific binding landscapes of individual TRs and eventually enable the over-time creation of a comprehensive reference library for TRex to identify vision-critical TRs at cellular resolution. Applying these innovative models to extensive ocular transcriptomic and epigenomic datasets, we will gain unprecedented insights into the genetic and regulatory basis of ocular biology at both tissue and cellular levels, accelerating the discovery of novel diagnostics and therapeutics. To maximize the impact of our research, TRex and sideTR will be made accessible to the research community through user-friendly software packages and a web platform, enabling software, data, resource sharing. We have assembled a highly collaborative, multi-disciplinary team with complementary expertise in all research areas needed for this project (e.g., Bayesian statistics, bioinformatics, single-cell sequencing, omics data analysis, ocular biology, and relevant clinical experience). Our team has meticulously curated 850 NGS datasets from various ocular tissue samples and cell lines under diverse physiological and pathological conditions, and ~16,000 TR ChIP-seq datasets documented in ~300 PubMed publications, with ~4000 more forthcoming. This rich collection of high-quality data ensures the viability and success of our study. Preliminary results have shown that the proposed methods compare favorably to the state-of-the-art methods for each task.
