wedap 1.3.0

WEDAP : weighted ensemble data analysis and plotting (pronounced we-dap)

wedap is primarily used to plot H5 files produced from running WESTPA.

mdap can be used to plot data files from analysis of standard MD simulations.

wekap can be used to plot flux values as rates from a WESTPA created direct.h5 file.

For a demo and summary of features, see this jupyter notebook.

Or view the same demo notebook on the documentation web page.

What's New in v1.3.0

This release fixes h5 files being left open and locked by wedap (e.g. after an error in a Jupyter notebook), which could block other programs like w_multi_west from writing to them (see the H5 file access section below), along with fixes to the H5save_out, succ_only, skip_basis and multiple h5 file options.

v1.2.0 replaced the previously removed/deprecated gooey-based GUI with a streamlit app (see the Web app (Streamlit) section below). See CHANGELOG.md for more info.

Requirements

• numpy
• matplotlib
• h5py
• tqdm
• gif

Installation

pip install wedap

Or install the latest development version from source:

git clone https://github.com/darianyang/wedap.git
cd wedap
pip install -e .

Examples

After installation, to run the CLI version and view available options:

wedap --help

Or:

wedap -h

To visualize the evolution of the pcoord for the example p53.h5 file via CLI:

wedap -h5 wedap/data/p53.h5

To do the same with the API:

import wedap
import matplotlib.pyplot as plt

wedap.H5_Plot(h5="wedap/data/p53.h5", data_type="evolution").plot()
plt.show()

The resulting p53.h5 file evolution plot will look like this:

See the examples directory for more realistic applications using the Python API.

Evolution plots are created by default using the CLI but average and instant probability distribution options are also available. To use one of your auxiliary datasets instead of the progress coordinate, just include the name of the aux dataset from your h5 file in the --Xname or --Yname fields:

wedap -h5 wedap/data/p53.h5 --data_type average --Xname dihedral_10 --Yname dihedral_11

Or:

wedap -h5 wedap/data/p53.h5 -dt average -X dihedral_10 -Y dihedral_11

The resulting p53.h5 file average plot of the dihedral aux datasets will look like this:

If you used a multi-dimensional progress coordinate and you want to use your pcoord for both the X and Y dimensions in a 2D average or instant plot, just use pcoord with the corresponding index set to the appropriate dimension (this also works with aux datasets which may have an additional dimension):

wedap -h5 wedap/data/p53.h5 --data_type average --Xname pcoord --Xindex 0 --Yname pcoord --Yindex 1

Or:

wedap -h5 wedap/data/p53.h5 -dt average -X pcoord -Xi 0 -Y pcoord -Yi 1

Or (since the default X options are the first pcoord, only the second pcoord needs to be specified):

wedap -h5 wedap/data/p53.h5 -dt average -Y pcoord -Yi 1

The resulting p53.h5 file average plot of the pcoord datasets will look like this:

H5 file access

wedap only keeps an h5 file open while one of its methods is running (e.g. pdist() or plot_trace()), and closes it again afterwards, even if there was an error. So an H5_Pdist or H5_Plot object left in a Jupyter notebook does not lock the file or block other programs like w_run or w_multi_west from writing to it. If you read from the h5 attribute of these objects directly, call close() when you are done.

When calling methods many times in a loop (e.g. get_parents()), use a with block to keep the file open for the whole loop instead of reopening it for each call, then it is closed at the end:

pdist = wedap.H5_Pdist(h5="west.h5", data_type="evolution")
with pdist:
    parents = [pdist.get_parents((10, seg)) for seg in range(50)]

Web app (Streamlit)

wedap ships with an optional browser-based interface built on Streamlit, providing a point-and-click alternative to the CLI (dataset dropdowns, live plot options, a downloadable figure, and a copy-pasteable equivalent Python snippet).

The app bundles all three tools as tabs (each tab keeps its own inputs when you switch between them):

• wedap — WESTPA west.h5 probability distributions and plots (1D/2D/3D, plot tracing, gif making).
• mdap — probability distributions and plots from standard MD analysis data files (.dat/.npy/...).
• wekap — WE kinetics: rates/MFPTs from a direct.h5 file (RED correction, multi-replicate averaging).

Install the optional web extra and launch it:

pip install "wedap[web]"
wedap-web

Or run it directly on the app module:

streamlit run wedap/web/app.py

This starts a local server (default port 8501) and opens the app in your browser at http://localhost:8501.

Running on a remote cluster

If you run wedap-web on a remote machine (e.g. an HPC login or compute node), the server runs there but your browser is local, so you need to forward the port over SSH.

First launch it headless on the cluster (optionally pick a port to avoid clashes):

wedap-web --server.headless true --server.port 8501

Then, from your local machine, open an SSH tunnel and browse to http://localhost:8501:

# from a login node:
ssh -L 8501:localhost:8501 user@cluster

# from a compute node (two-hop; cNODE is the node's hostname):
ssh -L 8501:cNODE:8501 user@cluster-login

Keep the SSH session open while you use the app. If port 8501 is already in use locally, forward to a different local port, e.g. -L 8600:localhost:8501, and browse http://localhost:8600.

Tip: For large west.h5 files, use the app's "Server path" input mode (point it at a path on the cluster) rather than the file uploader, which holds the file in memory.

Motivation

WESTPA already comes with some excellent analysis tools for generating probability distributions, so why is wedap needed?

wedap was originally built as a way to simplify the original WESTPA plotting pipeline:

Native WESTPA CLI-based Analysis Tools:

┌───────┐       w_pdist        ┌────────┐        plothist         ┌────────┐
│west.h5├─────────────────────►│pdist.h5├────────────────────────►│plot.pdf│
└───────┘ --construct-dataset  └────────┘ --postprocess-function  └────────┘
               module.py                      plot_settings.py

Analysis using wedap:

┌───────┐     wedap      ┌────────┐
│west.h5├───────────────►│plot.pdf│
└───────┘  CLI/Python    └────────┘

So wedap can generate plots with more flexibilty and less intermediate files, providing an especially useful way to plot aux datasets and explore your h5 file.

• The Python interface allows for advanced users to quickly generate a plot as a matplotlib axes object which can be further customized all in one Python script.
  • For example, the moviepy or gif package can be used with wedap to easily create a gif of your h5 file (see an example of this in wedap/h5_movie.py).
  • The actual data can also be easily extracted and then analyzed (see wedap/h5_cluster.py for an example of k-means clustering using the data from a WESTPA west.h5 file).
• A CLI is also available if using wedap on a system without access to a display.

Since the original implementation of wedap, many more features have been added that are not available using the WESTPA w_pdist and plothist tools, these include the following:

• Easy WE tracing and plotting by inputing an iteration and segment, or by inputing the X and Y value to then query and trace.
• 3D plots that replace the probability with another pcoord or aux dataset (plot_mode="scatter3d").
• Selective basis states (if you have multiple basis states, only plot the probability contributions from specific states).
  • See the skip_basis argument (available through the Python API only currently).
• More to come!

Note that the WESTPA analysis tools have features not available in wedap and may still be of interest to you.

Contributing

Have an idea for a feature to add to wedap? Let me know and I may be able to incorporate it (dty7@pitt.edu).

Or feel free to try developing it yourself! Features should be developed on branches. To create and switch to a branch, use the command:

git checkout -b new_branch_name

To switch to an existing branch, use:

git checkout branch_name

To submit your feature to be incorporated into the main branch, you should submit a Pull Request. The repository maintainers will review your pull request before accepting your changes.

Copyright

Copyright (c) 2021-2026, Darian Yang