Welcome to Use Xponge!

Introduction

Xponge is a lightweight and easy to customize python package to perform pre- and post-processing of molecular simulations.

What can Xponge do?

Xponge includes three major categories of functionality, namely, the simulation system construction, simulation data transformation and analysis, and automated workflows for complex simulations. Xponge is mainly designed for the molecular dynamics (MD) program SPONGE[1], but it can also output some general format files such as mol2 and PDB, so it may help the other molecular modelling programs too.

Installation

Xponge can be used on all operating systems (Windows/Linux/macOS), including its quantum chemistry workflows. Xponge prefers PySCF and automatically falls back to Psi4 when PySCF is not installed; these workflows report an error if neither backend is available. Until official Windows PySCF wheels are released, build and install a compatible PySCF wheel separately to use the PySCF backend.

1. pip install

pip install Xponge

2. source setup

2.1 Download or clone the source from the gitee or github repository

The gitee repository is here. The github repository is here.
```
  git clone http://gitee.com/gao_hyp_xyj_admin/xponge.git
  git clone http://github.com/xia-yijie/xponge.git
```
2.2 Open the directory where you download or clone the repository
2.3 (Optional) Configure the environment

It is recommended to use conda to configure the environment. Two yml files named install_requirements.yml and extras_requirements.yml are provided in the repository.

It is recommanded to use the file install_requirements.yml to configure the environment. The file will only install the basic dependent packages. If a ModuleNotFoundError is raised when you are using Xponge, then install the module. This allows you to avoid installing many modules that you will never use, and also makes Xponge more cross-platform compatible. Here are the commands to use install_requirements.yml.
```
conda env create -f install_requirements.yml
conda activate Xponge
```
All the dependent packages are listed in the dependencies section. If you don't want to install the dependent packages one by one (which can be really annoying), the file extras_requirements.yml can help you with the environment configuration except the packages mindspore and mindsponge. The two packages should be installed according to your device (e.g. whether the backend is CPU, GPU or Huawei Ascend Processors) and can not be simply installed by conda. Here are the commands to use extras_requirements.yml.
```
conda env create -f extras_requirements.yml
conda activate Xponge
```
2.4 Run the command
```
python setup.py install
```

Installation check

There are some unit tests in Xponge. You can do the basic test to check whether the installation is successful like this:

Xponge test --do base -o test --verbose 1

Here, Xponge can be replaced to python -m Xponge, python3 -m Xponge and so on according to your settings of the environmental variables. Some files will be generated after the test is finished.

Quickstart

Here is a simple example.

import Xponge
# Import the force field you need
import Xponge.forcefield.amber.ff14sb
# Build the molecule like this
peptide = ACE + ALA + NME
# or like this
peptide2 = NALA + ALA * 10 + CALA
# or like this
peptide3 = Xponge.Get_Peptide_From_Sequence("AAAAA")
# See the documentation for more usage!
# Save them as your favorite format
Xponge.save_pdb(peptide, "ala.pdb")
Xponge.save_mol2(peptide2, "ala12.mol2")
Xponge.save_sponge_input(peptide3, "ala5")

Metal force-field overlays

Xponge.metal_assignment extends an already loaded and normally assigned Molecule; it is not a separate structure-building pipeline. Callers provide the locked metal sites, topology mapping, electronic state and parameter protocol, then continue with the ordinary saver:

import Xponge
import Xponge.forcefield.amber.ff19sb
import Xponge.forcefield.amber.tip3p

mol = Xponge.load_mmcif("system.cif")
result = Xponge.metal_assignment.assign(
    mol,
    package=metal_assignment_package,
    atom_mapping=atom_mapping,
    operation="fit-bonded",
    water_model="tip3p",
    bonded_fit_input=bonded_fit_input,
)
Xponge.save_sponge_input(result.molecule, "system", format="bundle")

The API also exposes separate parameterize and transactional apply operations. Xponge never splits residues, chooses coordination bonds or builds caps; the caller must provide a complete topology and stable mapping.

For manual_bonded, bonded_fit_input carries explicit, hash-closed reference bond lengths and angles plus site force constants. Xponge validates term coverage and applies the resulting overlay without running Hessian or Seminario and without inferring terms from distances. The potential convention is E = k * delta^2; canonical units are kcal/mol/angstrom^2 for bonds and kcal/mol/rad^2 for angles. This path is intended for experimental engineering validation, not as a claim of QM-fitted scientific accuracy.

Amber lipid force fields

Lipid17 and Lipid21 are available as mutually exclusive base lipid force fields:

import Xponge.forcefield.amber.ff14sb   # optional protein family
import Xponge.forcefield.amber.gaff2    # optional small-molecule family
import Xponge.forcefield.amber.lipid21  # or lipid17, but not both

Loading either lipid force field also loads the bundled PACKMOL-Memgen-derived PI/phosphoinositide/LysoPL extension. Lipids use Amber's split residue templates (for example PA + SPM + SA); automatic conversion of full names such as PSM or POPC is not provided. The extension combines documented Lipid, GLYCAM, phosphate, and selected GAFF2-derived terms; its references are printed on load.

The protein (ff14sb/ff19sb), small_molecule (gaff/gaff2), and lipid (lipid17/lipid21) families each allow one active base force field per Python process. Use a fresh process to compare alternatives.

Then we can see ala12.mol2 in VMD:

Here is another simple example.

import Xponge
import Xponge.forcefield.amber.tip3p

box = Xponge.BlockRegion(0, 0, 0, 60, 60, 60)
region_1 = Xponge.BlockRegion(0, 0, 20, 20, 20, 40)
region_2 = Xponge.BlockRegion(0, 0, 40, 20, 20, 60)
region_3 = Xponge.BlockRegion(0, 0, 0, 20, 20, 20)
region_4 = Xponge.SphereRegion(20, 10, 30, 10)
region_5 = Xponge.BlockRegion(20, 0, 20, 60, 60, 60)
region_2or3 = Xponge.UnionRegion(region_2, region_3)
region_4and5 = Xponge.IntersectRegion(region_4, region_5)
t = Xponge.Lattice("bcc", basis_molecule=CL, scale=4)
t2 = Xponge.Lattice("fcc", basis_molecule=K, scale=3)
t3 = Xponge.Lattice("sc", basis_molecule=NA, scale=3)
mol = t.Create(box, region_1)
mol = t2.create(box, region_2or3, mol)
mol = t3.create(box, region_4and5, mol)
Xponge.Save_PDB(mol, "out.pdb")

Then we can see out.pdb in VMD:

Detailed usage and API documentation

All can be seen here.

SPONGE bundled inputs

XPONGE can write the topology, protocol, and structural restart HDF5 inputs directly from a Molecule, Residue, or ResidueType:

import Xponge
import Xponge.forcefield.amber.ff14sb

peptide = Xponge.get_peptide_from_sequence("AAAAA")
Xponge.save_sponge_input(
    peptide,
    prefix="ala5",
    dirname="inputs",
    format="bundle",
)

This creates:

inputs/ala5_topology.spgt.h5
inputs/ala5_protocol.spgp.h5
inputs/ala5_restart.spgr.h5

Bind them in the SPONGE mdin with input_h5_topology_path, input_h5_protocol_path, and input_h5_restart_path. The saver does not add run policy such as mode, step limit, thermostat, or output paths.

Reusable simulation protocol objects can be attached without compiling legacy cv.toml or other text inputs. Object references use stable names and are validated before any artifact is published:

protocol = Xponge.SpongeProtocol(
    collective_variables=(
        Xponge.ProtocolCollectiveVariable(
            name="distance",
            type="distance",
            atom_indices=(0, 1),
            sigma=(0.5,),
        ),
    ),
    cv_restraints=(
        Xponge.ProtocolCVRestraint(
            name="umbrella",
            cv_refs=("distance",),
            weight=(1.0,),
            reference=(1.5,),
        ),
    ),
    metadynamics=(
        Xponge.ProtocolMetadynamics(
            name="bias",
            cv_refs=("distance",),
            grid_min=(0.0,),
            grid_max=(10.0,),
            grid_count=(64,),
        ),
    ),
    steering=Xponge.ProtocolSteering(
        cv_refs=("distance",),
        weight=(0.25,),
    ),
    hard_wall=Xponge.ProtocolHardWall(
        bounds_low=(0.0, None, None),
        bounds_high=(40.0, None, None),
    ),
)

Xponge.save_sponge_input(
    peptide,
    prefix="ala5",
    dirname="inputs",
    format="bundle",
    protocol=protocol,
)

The native object API currently covers scalar CVs, extra distance constraints, positional restraints, CV harmonic restraints, one enabled metadynamics object, one static steering definition, and one typed SITS protocol. SITS method and atom-selection fields are written to protocol.spgp.h5; its versioned launch state (nk, optional log_norm, and optional log_nk) is written to restart.spgr.h5. Named ProtocolSoftWall objects are stored as typed columnar definitions and are compiled directly by SPONGE without an intermediate configuration file. ProtocolHardWall uses None for an unbounded axis, requires at least one finite bound, and defaults to rejecting NPT unless allow_npt=True explicitly opts in. RMSD CV reference_coordinates are written inline to /cv/<name>/coordinate in protocol.spgp.h5, with one finite XYZ row per selected atom, in atom_indices or atom_refs order. SPONGE also accepts the old restart reference path; if both are present, they must agree. Positional reference coordinates remain in the structural restart artifact. A positional-restraint reference contains one XYZ row for every system atom; atom_indices selects which rows are restrained.

SITS is represented independently rather than compiled from a legacy section:

sits_protocol = Xponge.SpongeProtocol(
    sits=Xponge.ProtocolSITS(
        mode="production",
        atom_indices=(0, 1),
        temperature_ladder=(280.0, 320.0),
        initial_nk=(1.0, 4.0),
    )
)

The default remains the legacy raw-text format. It can be selected explicitly with format="raw"; the raw implementation is also available directly as Xponge.save_sponge_input_raw. Xponge.save_sponge_input_bundle remains available for callers that prefer the format-specific API.

Existing direct/legacy input cases can be converted in either direction:

Xponge legacy-to-bundle CASE_DIR -m mdin.spg.toml -o CONVERTED
Xponge bundle-to-legacy CONVERTED/bundle -o RESTORED --prefix system

bundle-to-legacy also exports native CV definitions and virtual atoms. RMSD reference coordinates are included directly as coordinate = ... in the generated CV file, preserving atom order and avoiding external reference files. Both inline protocol references and the old restart reference path are accepted; conflicting references or malformed coordinates fail before files are written. Disabled native objects are omitted. Legacy-to-bundle conversion retains these self-contained CV sections under /cv/config, so subsequent exports preserve their values.

Reverse conversion treats typed HDF5 datasets as authoritative. Embedded legacy sidecars are used only for contracts without a typed representation. The current reader accepts the xponge.legacy_to_bundle.v1 input schema and rejects unknown schema names or versions in strict mode. Use --dry-run to validate and inspect the conversion manifest without writing output files. Existing targets are preserved unless --overwrite is specified.

Bundled topology and trajectory files are registered as MDAnalysis formats. Import the XPONGE adapter once, then use the standard Universe entry point; no mdin, protocol, or restart file is required for analysis:

import MDAnalysis as mda
import Xponge.analysis.md_analysis

universe = mda.Universe(
    "topology.spgt.h5",
    "trajectory.spg.h5md",
    topology_format="SPONGE_TOPOLOGY_H5",
    format="SPONGE_H5MD",
    particle_stream="all",
)

Xponge.analysis.md_analysis.load_bundle_universe(topology, trajectory) is a stricter convenience wrapper that additionally verifies topology and atom order hashes. The same SPONGE_H5MD reader retains support for historical /particles/trajectory and numbered walker layouts through the walker keyword, and delegates ordinary third-party H5MD files to MDAnalysis' native reader.

CLI quick start (trajectory analysis)

The Xponge traj subcommand provides cpptraj-like post-analysis for SPONGE trajectories.

Basic syntax:

Xponge traj -p TOPO -c TRAJ [-c TRAJ ...] -b BOX [-b BOX ...] [--traj-mode {separate,concat}] -o OUTDIR <analysis> [options]

Notes:

-c/--traj and -b/--box are repeatable; use -b once for a single box file or repeat it to match each trajectory.
--traj-mode separate means multiple -c inputs are analyzed independently, with outputs written to per-trajectory subfolders.
--traj-mode concat means multiple -c inputs are treated as successive segments of one trajectory and analyzed together in the given order.
In -i/--input command files, traj=/box= accept comma-separated lists.
In -i/--input command files, traj_mode= accepts separate or concat and defaults to the command-line --traj-mode value.

Common analyses:

# RMSD
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out rmsd -s "backbone" --dt-ps 1

# RMSF
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out rmsf -s "backbone and name CA"

# Radius of gyration
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out rgyr -s "protein and backbone" --dt-ps 1

# Hydrogen bonds
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out hbond --between "protein,resname SOL" --dt-ps 1

# PCA
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out pca -n 2 -s "backbone"

# Free energy surface (CVs: rmsd:SELECTION, rgyr:SELECTION)
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out fes --cv1 "rmsd:backbone" --cv2 "rgyr:protein and backbone" --bins 20 --temperature 300

# Extract PDBs at specific times (ns)
Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out extract_pdb --times 0 0.1 0.2 --dt-ps 1 -s "all"

# multiple trajectories analyzed separately
Xponge traj -p top.txt -c mdcrd_1.dat -c mdcrd_2.dat -b mdbox_1.txt -b mdbox_2.txt -o out rmsd -s "backbone" --dt-ps 1 --traj-mode separate

# successive trajectory segments concatenated into one analysis
Xponge traj -p top.txt -c md_0_100.dat -c md_100_200.dat -c md_200_300.dat -b box_0_100.txt -b box_100_200.txt -b box_200_300.txt -o out rmsd -s "backbone" --dt-ps 1 --traj-mode concat

cpptraj-like command file:

# analysis.in
rmsd selection="backbone" dt_ps=1
rmsf selection="backbone and name CA"
rgyr selection="protein and backbone" dt_ps=1
hbond between="protein,resname SOL" dt_ps=1
pca n=2 selection="backbone"
fes cv1="rmsd:backbone" cv2="rgyr:protein and backbone" bins=20 temp=300
extract_pdb times=0,0.1,0.2 selection="all" dt_ps=1

# Optional per-line trajectory/box override (comma-separated for multiple)
rmsd selection="backbone" dt_ps=1 traj="mdcrd_1.dat,mdcrd_2.dat" box="mdbox_1.txt,mdbox_2.txt" traj_mode=concat

Xponge traj -p top.txt -c mdcrd.dat -b mdbox.txt -o out -i analysis.in

Outputs are written as PNG figures and JSON data files in OUTDIR.

Convert JSON outputs to CSV for tools such as gnuplot:

# convert one analysis result
Xponge json2csv -i out/rmsd_xxxxxx.json

# choose the output CSV path explicitly
Xponge json2csv -i out/free_energy_surface.json -o out/free_energy_surface.csv

# convert every JSON file in a result directory
Xponge json2csv -i out -o out_csv -r

Supported JSON schemas currently include line-series results with x/y, extract_pdb time-file tables, free-energy surfaces, PCA outputs, and extra sidecar CSVs for fields such as RMSD statistics or raw hydrogen-bond lists.

Contribution Guideline

If you want to contribute to the main codebase or report some issues, see here for the guides.

Dependencies

Xponge does not depend on other packages except numpy for its basic use.

However, there are some complicated functions that depend on some other packages. If you do not install the dependent package, you can not use the related functions.

Here is the list of all packages which may be uesd:

package name	description	how to install
XpongeLib	c/c++ compiled library for Xponge	`pip install XpongeLib`
pyscf [2-4]	quantum chemistry	`pip install pyscf`
geometric[5]	geometry optimization	`pip install geometric`
rdkit[6]	cheminformatics	`pip install rdkit`
MDAnalysis[7-8]	trajectory analysis	`pip install MDAnalysis`
matplotlib	plot and visualization	`pip install matplotlib`
mindspore[9]	AI framework for machine learning	See the official website
mindsponge[1]	end-to-end differentiable MD	See the official website

References

[0] Y. Xia, Y. Q. Gao, J. Open Source Softw. (2022) DOI:10.21105/joss.04467

[1] Y.-P. Huang, et al. Chinese J. Chem. (2022) DOI: 10.1002/cjoc.202100456

[2] Q. Sun, et al. J. Chem. Phys. (2020) DOI: 10.1063/5.0006074

[3] Q. Sun, et al. Wiley Interdiscip. Rev. Comput. Mol. Sci. (2018) DOI: 10.1002/wcms.1340

[4] Q. Sun, J. Comp. Chem. (2015) DOI: 10.1002/jcc.23981

[5] L.-P. Wang, C.C. Song, J. Chem. Phys. (2016) DOI: 10.1063/1.4952956

[6] RDKit: Open-source cheminformatics. https://www.rdkit.org

[7] R. J. Gowers, et al. Proceedings of the 15th Python in Science Conference (2016) DOI: 10.25080/majora-629e541a-00e

[8] N. Michaud-Agrawal, et al. J. Comput. Chem. (2011) DOI: 10.1002/jcc.21787

[9] MindSpore: An Open AI Framwork. https://www.mindspore.cn/

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
.serena		.serena
README_PICTURE		README_PICTURE
Xponge		Xponge
docs		docs
packaging/mokda-xponge		packaging/mokda-xponge
scripts		scripts
tests		tests
tools		tools
.gitignore		.gitignore
.python-version		.python-version
LICENSE		LICENSE
README.md		README.md
contribution_guide.rst		contribution_guide.rst
extras_requirements.yml		extras_requirements.yml
install_requirements.yml		install_requirements.yml
main.py		main.py
paper.bib		paper.bib
paper.md		paper.md
pylintrc		pylintrc
pyproject.toml		pyproject.toml
setup.py		setup.py
uv.lock		uv.lock

Navigation Menu

Repository files navigation

Welcome to Use Xponge!

Introduction

What can Xponge do?

Installation

1. pip install

2. source setup

Installation check

Quickstart

Metal force-field overlays

Amber lipid force fields

Detailed usage and API documentation

SPONGE bundled inputs

CLI quick start (trajectory analysis)

Contribution Guideline

Dependencies

References

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Welcome to Use Xponge!

Introduction

What can Xponge do?

Installation

1. pip install

2. source setup

Installation check

Quickstart

Metal force-field overlays

Amber lipid force fields

Detailed usage and API documentation

SPONGE bundled inputs

CLI quick start (trajectory analysis)

Contribution Guideline

Dependencies

References

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages