An extensive reference atlas of cell type specific gene regulatory interactions covering 15 organ systems and more than 500 cell types has been created by processing large integrated cell atlas datasets from 12 publications. This gene regulatory atlas has been prepared as a database, CellRegulon DB, including a web-interface and python package for programmatic access. It can be used together with single cell or bulk query data to obtain predictions of which transcription factors in which cell types and tissues explain observed differences in gene expression. In addition, the regulon atlas itself was analysed to examine the activity of TFs across cell types and TF co-regulation. Querying the database has been demonstrated at the example of disease gene signatures (adult- and childhood-onset asthma), a single cell multiome lung cell atlas and data from TF overexpression experiments aiming to differentiate iPSCs into more specialised cell types.
As a new way to prioritise connections in a reference network, an approach based on network propagation has been developed. Heterogeneous, cell type specific data is passed through the network and significantly enriched nodes are identified. These could represent active transcription factors that drive gene expression and chromatin accessibility in a given cell type. To increase the scope of the problem, genomic scores derived from GWAS summary statistics have been tested as another type of input data. Mutations in complex diseases often affect enhancers. Therefore, when mapping summary statistics based scores onto the network and integrating it with cell type specific data, TFs which are likely to be functionally affected by those mutations can therefore be identified in a cell type specific manner. This approach has been developed and applied as part of a third, data analysis based line-of-work described next.
A new dataset of first trimester human skeletal development has been extensively analysed with a focus on gene regulation along cell differentiation trajectories and across anatomic locations. The dataset consists of more than 300k cell nuclei profiled with both RNA and ATAC sequencing and spans various time points between 5-11 pcw across 5 locations. Leveraging both modalities, enhancer-GRNs have been predicted for developmental trajectories including osteogenesis and chondrogenesis. Changes of TF activity were analysed and effects of TF perturbations were predicted, in particular for mutations known to cause craniosynostosis, a genetic condition with premature fusion of bone plates in the skull. In addition, GWAS summary statistics for hip osteoarthritis have been integrated with single cell data in a newly developed approach. Interestingly, enrichments of TFs involved in bone formation across osteogenic cell types were observed, pointing to a role in hip shape formation, which when altered may lead to disease later in life.
The results generated as part of this project were planned to be published in one or more scientific papers. Changes in the project plan due to new scientific developments and an increased scope of the work led to delays in the dissemination of the results. However, a manuscript covering the application of GRN inference to skeletal development is currently under revision, as well as a second manuscript combining GWAS summary statistics with single cell data. Finally, a manuscript about CellRegulon DB, a database of cell type specific reference-GRNs is going to be submitted soon. The project results have been and will be further disseminated in additional scientific meetings and through planned outreach activities.
