DTNE - Diffusive Topology Neighbor Embedding

DTNE (Diffusive Topology Neighbor Embedding) is a Python tool that implements a novel manifold learning framework. By leveraging a diffusive process, DTNE constructs a manifold distance matrix, enabling key analyses such as dimensionality reduction, pseudotime ordering, and cluster identification, providing valuable insights into complex datasets.

Features

• Dimensionality Reduction: By preserving manifold geodesic distances, DTNE provides accurate low-dimensional projections of high-dimensional single-cell data.
• Pseudotime Inference: Infer developmental trajectories by leveraging the manifold distance matrix, improving the identification of lineage progression.
• Clustering: DTNE enables clustering by utilizing the manifold distance matrix, which captures both local and global data structures.

Installation

install the package DTNE by running the following command in the terminal: pip install .

Quick Start

DTNE requires input data in the form of a high-dimensional matrix. Suppose you have loaded a single-cell data X in Python.

1. Dimensionality Reduction

   DTNE can be used for dimensionality reduction:

   from dtne import *
   dtne_operator = DTNE(k_neighbors = 10,l=2) 
   Y = dtne_operator.fit_transform(X)
   

2. Pseudotime Inference

   DTNE allows for pseudotime inference to analyze cell development trajectories:

   dtne_pseudotime = dtne_operator.order_cells(root_cells=[0])
   

3. Clustering DTNE supports clustering based on the computed manifold distances:

   dtne_cluster = dtne_operator.cluster_cells(n_clusters=8)
   

Tutorial and Example Notebooks

Several example Jupyter notebooks are provided in the notebooks/ directory, demonstrating DTNE’s usage on various single-cell datasets.

Citation

If you use DTNE in your research, please cite the paper.

License

DTNE is licensed under the MIT License.