DyCaP: Identifying recurrent dynamic correlation patterns in single-cell RNA-seq data

Background:

Many existing methods for single-cell RNA-seq analysis treat gene-gene correlations as static quantities evaluated at a single time point, while attempts to model correlations themselves as functions that vary over time have been limited.

Methods:

Here, we propose a new method, DyCaP, which defines gene-gene correlations along pseudotime as functions and compares correlation structures based on the similarity of their temporal patterns.

Results:

Applying DyCaP to single-cell RNA-seq data, we identified recurrent dynamic correlation patterns characterized by temporal trajectories exhibiting similar shapes, which are not captured by conventional static correlation analyses.

Conclusions:

DyCaP provides a new framework for hypothesis generation based on the temporal structure of gene-gene correlations.

GitHub Repository: https://github.com/aoyamaforresearch/DyCaP