PHATE (Potential of Heat-diffusion for Affinity-based Trajectory Embedding)

To install the Python version of PHATE from the GitHub repository, clone the repository recursively and run the setup script from the Python directory.

git clone --recursive git://github.com/KrishnaswamyLab/PHATE.git
cd Python
python setup.py install --user

The easiest way to install the Python version of PHATE is using pip.

pip install --user phate

To use PHATE in R, you must first install the Python package (pip install --user phate) and then install the phateR package from CRAN.

install.packages("phateR")

Using PHATE in R requires the Python phate package to be installed first.

Option 1: Standard Installation

1. Install the Python package via terminal:

   pip install --user phate

2. Install phateR from CRAN in R:

   install.packages("phateR")

Option 2: Development Installation (using devtools and reticulate) This installs the development version directly from GitHub:

if (!suppressWarnings(require(devtools))) install.packages("devtools")
reticulate::py_install("phate", pip=TRUE)
devtools::install_github("KrishnaswamyLab/phateR")

You can install the development version of phateR directly from GitHub using devtools.

if (!suppressWarnings(require(devtools))) install.packages("devtools")
devtools::install_github("KrishnaswamyLab/phateR")

The primary entry point for performing PHATE dimensionality reduction in MATLAB is the phate.m function. To access the built-in documentation and usage instructions within the MATLAB command window, use the help command.

help PHATE

You can install the Python version of PHATE using pip. All dependencies are installed automatically.

pip install --user phate

To perform PHATE dimensionality reduction on a dataset in Python, initialize a phate.PHATE() operator and call fit_transform(data).

Input Data Requirements:

• The data matrix should have cells (observations) on rows and genes (features) on columns.
• Supported data types: numpy.array, scipy.spmatrix, pandas.DataFrame, and anndata.AnnData.

import phate
phate_op = phate.PHATE()
data_phate = phate_op.fit_transform(data)

To install the Python version from the GitHub repository, clone the repository recursively and run the setup script.

git clone --recursive git://github.com/KrishnaswamyLab/PHATE.git
cd PHATE/Python
python setup.py install --user

The main entry point for performing PHATE dimensionality reduction in MATLAB is the phate.m function. To access the built-in documentation and usage instructions within the MATLAB command window, use the help command.

help PHATE

To perform PHATE dimensionality reduction, instantiate a phate.PHATE operator, then use the fit_transform method on your input data. You can adjust parameters using set_params. After fitting, you can use transform() to project data into a different number of components (e.g., moving from 2D to 3D).

import phate
import scprep

# Generate or load your data
tree_data, tree_clusters = phate.tree.gen_dla()

# Initialize the PHATE operator
# k: number of neighbors
# t: diffusion time
phate_operator = phate.PHATE(k=15, t=100)

# Fit and transform the data to 2D
tree_phate = phate_operator.fit_transform(tree_data)
scprep.plot.scatter2d(tree_phate, c=tree_clusters)

# Change dimensionality to 3D
phate_operator.set_params(n_components=3)
tree_phate = phate_operator.transform()
scprep.plot.rotate_scatter3d(tree_phate, c=tree_clusters)

Detailed usage examples and visualizations are available in Jupyter notebooks:

• Single cell RNA-seq data: Demonstrates PHATE usage and visualization for biological datasets.

  • View on nbviewer
  • Open in Google Colab

• Artificial tree data: A detailed walkthrough using the phate.tree.gen_dla() dataset.

  • View on nbviewer
  • Open in Google Colab