Introduction

For a quick intro have a look at the following video:

You can find more information in these two articles:

Lemke, Tobias, and Christine Peter. "EncoderMap: Dimensionality Reduction and Generation of Molecule Conformations." Journal of chemical theory and computation 15.2 (2019): 1209-1215.

Lemke, T., Berg, A., Jain, A., & Peter, C. "EncoderMap (II): Visualizing important molecular motions with improved generation of protein conformations" Journal of chemical information and modeling (2019).

Documentation

Find the documentation at:

https://ag-peter.github.io/encodermap/

Quick start

To get started right away, you can try out the tutorial notebooks at mybinder.org:

Installation

Version 3.0.0 is currently not on PyPI. Install with:

pip install git+https://github.com/AG-Peter/encodermap.git

Install optional packages for using MD data:

pip install -r https://raw.githubusercontent.com/AG-Peter/encodermap/main/md_requirements.txt

EncoderMap requires TensorFlow >=2.9 to be installed. Follow the instructions on the TensorFlow website to install it either in the cpu or gpu version. Then install EncoderMap with pip. If you want to install it in your home directory use:

pip3 install --user encodermap

If you are in a virtual environment use:

pip3 install encodermap

Minimal Example

This example shows how to use EncoderMap to project points from a high dimensional data set to a low dimensional space using the default parameters. In the data set, each row should represent one data point and the number of columns should be equal to the number of dimensions.

import encodermap as em
import numpy as np

high_dimensional_data = np.loadtxt("my_high_d_data.csv", delimiter=",")
parameters = em.Parameters()

e_map = em.EncoderMap(parameters, high_dimensional_data)
e_map.train()

low_dimensional_projection = e_map.encode(high_dimensional_data)

The resulting low_dimensional_projection array has the same number of rows as the high_dimensional_data but the number of columns is two as high dimensional points are projected to a 2d space with default settings.

In contrast to many other dimensionality reduction algorithms EncoderMap does not only allow to efficiently project form a high dimensional to a low dimensional space. Also the generation of new high dimensional points for any given points in the low dimensional space is possible:

low_d_points = np.array([[0.1, 0.2], [0.3, 0.4], [0.2, 0.1]])
newly_generated_high_d_points = e_map.generate(low_d_points)

Tutorials

To get started please check out the tutorials.

Also, have a look at the examples.

Documentation

More information is available in the documentations.

Questions

If you have any questions you can have a look at the FAQ (not very extensive yet), and you are most welcome to open an issue here on GitHub.

Citations

@article{lemke2019encodermap,
  title={Encodermap: Dimensionality reduction and generation of molecule conformations},
  author={Lemke, Tobias and Peter, Christine},
  journal={Journal of chemical theory and computation},
  volume={15},
  number={2},
  pages={1209--1215},
  year={2019},
  publisher={ACS Publications}
}

@article{lemke2019encodermap,
  title={EncoderMap (II): Visualizing important molecular motions with improved generation of protein conformations},
  author={Lemke, Tobias and Berg, Andrej and Jain, Alok and Peter, Christine},
  journal={Journal of Chemical Information and Modeling},
  volume={59},
  number={11},
  pages={4550--4560},
  year={2019},
  publisher={ACS Publications}
}