Cell-cell interactions mediated by ligand-receptor complexes are critical to coordinating organismal development and functions. Majority of studies focus on the key time point, the mediator cell types or the critical genes during organismal development or diseases. However, most existing methods are specifically designed for stationary paired samples, there hasn't been a repository to deal with inferring cell-cell communications in developmental series RNA-seq data, which usually contains multiple developmental stages. Here we present a cell-cell interaction networks inference method and its application for developmental series RNA-seq data (termed dsCellNet) from the developing and aging brain. dsCellNet is implemented as an open-access R package on GitHub. It identifies interactions according to protein localizations and reinforces them via dynamic time warping within each time point and between adjacent time points, respectively. Then, fuzzy clustering of those interactions helps us refine key time points and connections. Compared to other published methods, our methods display high accuracy and high tolerance based on both simulated data and real data. Moreover, our methods can help identify the most active cell type and genes, which may provide a powerful tool to uncover the mechanisms underlying the organismal development and disease.
Keywords: Cell-cell communications; Computational methods; Developmental series analysis; Single-cell RNA sequencing.
© 2022 The Authors.