Welcome to DEPF’s documentation!

Date: March 15, 2023. Version: 1.0.0

Contact: Thank you for using DEPF! Any questions, suggestions or advices are welcome. Email address: fanyi21@mails.jlu.edu.cn, lixt314@jlu.edu.cn.

paper: Reliable Identification and Interpretation of Single-cell Molecular Heterogeneity and Transcriptional Regulation using Dynamic Ensemble Pruning

Overview:

DEPF framework

DEPF is constructed based on four modules (Normalization, Hierarchical Autoencoder, Clustering Ensemble, Dynamic Ensemble Pruning) developed by ourselves.

  • Normalization: The expression data are rescaled to a range of 0 to 1 for each cell.

  • Hierarchical Autoencoder: The normalized data are mapped to multiple low-dimensional latent spaces.

  • Clustering Ensemble: An effective basic clustering algorithm is employed to address the non-linear embedding in the latent space to produce multiple underlying cluster results to generate cluster ensemble.

  • Dynamic Ensemble Pruning: The low-quality basic clusterings in the ensemble are dynamically pruned away.