Statistical Bioinformatics Webinar: Kamimoto -- CellOracle: Dissecting cell identity via network inference and in silico gene perturbation

Presented by Dr.  Kenji Kamimoto (Washington University in Saint Louis) 

Recent technological advances in single-cell sequencing enable the acquisition of
multi-dimensional data in a high-throughput manner.  These technologies reveal the
existence of heterogeneity and the diversity of cell states and identities.  To reveal
the regulatory mechanism underlying such phenomena, many computational Gene Regulatory
Network (GRN) inference methods have been proposed.  However, understanding biological
events from a GRN perspective remains difficult.  Even if a computational algorithm can
infer GRN, the biological network is so complex that it is challenging to understand how
it systematically dictates cell identities.  There is significant demand for new
methodologies that bridge the gap between cellular phenotypes and the underlying GRN.
Thus, we have developed a new method, CellOracle, a new computational approach for the
inference and analysis of GRN.  By utilizing machine learning algorithms and genetic
information, CellOracle infers sample-specific GRN configurations from single-cell
RNA-seq and ATAC-seq data.  Our GRN models are designed to be used for the simulation of
cell identity changes in response to gene perturbation.  This simulation enables network
configurations to be interrogated in relation to cell-fate regulation, facilitating
their interpretation.  To validate CellOracle’s GRN inference method, we present
benchmarking on various tissues and cell-types.  We also validate the efficacy of
CellOracle to recapitulate known outcomes of well-characterized gene perturbations in
developmental processes, including mouse hematopoiesis and zebrafish embryogenesis.  Our
benchmarking and validation results demonstrate the efficacy of CellOracle to infer and
interpret the dynamics of GRN configurations, promoting new mechanistic insights into
the regulation of cell identity.