Executive Summary : | Network principles and approaches are increasingly being applied to big data analytics, particularly in genomics, which is the leading contributor of big data to scientific research. Single-cell sequencing technology enables experimental biologists to generate gene expression data at the resolution of individual cells, generating millions of transcripts over millions of cells in a single experiment in a short span of time. However, this data requires novel statistical methods and tools to extract relevant biological information to understand complex biological processes like cancer, embryogenesis, and abiotic stresses in crops. Researchers have started importing methods from bulk RNA-seq to analyze scRNA-seq data, but these methods may not be efficient for handling the special features of single-cell data, such as cellular heterogeneity, zero inflation, excess overdispersion, and lower transcriptional capture efficiency. To address this, statisticians and computational biologists have developed novel gene network methods and tools exclusively for scRNA-seq data analytics. Existing approaches are based on correlation, Boolean, or ordinary linear differential equation techniques, which are simple and easy to interpret but often lead to spurious interactions and fail to tell the directions of interactions, crucial to the cell's unique molecular function. This project proposes developing statistical methods and AI-based tools for various aspects of gene network analysis for scRNA-seq data, focusing on zero-inflated models under a Bayesian computational framework and resampling-based statistical sound frameworks. User-friendly online and offline tools will be developed for stakeholders to benefit from these developments. |