This repository contains molecular dynamics (MD) simulations and trajectory analyses of the membrane-embedded mannosyltransferase ALG9.
The goal of this project is to understand the donor selectivity and key interactions of ALG9 during glycosylation.

This study aims to answer three main questions:

1. What does the mannose donor binding pose look like in the absence of experimental donor data?
2. What determines ALG9’s selectivity toward the mannose donor compared to glucose or 2-fluoro-mannose donors?
3. Which ALG9 residues are critical for donor stabilization, and why?

To address these:

• (a) Simulations were performed starting from the cryo-EM structure of ALG9 with a speculated mannose donor binding pose.
• (b) Simulations with glucose and 2-fluoro-mannose donors were conducted to assess donor selectivity.
• (c) Mutations N288A, H366A, and D82A were simulated to probe the role of key residues in donor stabilization.

Example MD simulation input files are provided in the folder MDsimulations/.
For quick testing, the demonstration protocol uses 50× fewer steps to allow short test runs.

You can run the short test directly by executing:

bash MD.sh

This script performs energy minimization, equilibration, and production in sequence.

To perform full-length production runs, replace the demonstration input files with the ones in
MDsimulations/longer_input_files/.

A full list of Python dependencies is provided in environment.yml.

All trajectory analysis is performed in Python (Jupyter Notebook) using:

• numpy and pytraj for data processing
• matplotlib and seaborn for visualization

The analysis scripts automatically compute:

• RMSD
• RMSF
• Interatomic distances
• Hydrogen bonds

By specifying the folder containing topology and trajectory files, all analyses can be executed with:

obj.run_analyze()

Upon completion:

• Figures are saved in the Figures/ directory.
• Analyzed objects (in .pkl format) are saved in the objects/ directory for faster future access.

The .pkl files contain precomputed analysis results, allowing users to regenerate figures without reloading trajectories.

All plots included in the related publication are automatically produced during analysis.

ALG9_project/
│
├── MDsimulations/
│   ├── input_files/                # Example minimized test inputs
│   ├── longer_input_files/         # Full-length production inputs
│   └── MD.sh                       # Run script for test simulations
│
├── analysis/
│   ├── ALG9_initial.ipynb          # Jupyter notebook for the analysis on the initial runs
│                                     with a guessed donor binding pose
│   ├── ALG9.ipynb                  # Jupyter notebook for the analysis on the later runs 
|                                     from the binding pose revealed from the initial runs
│   ├── objects/                    # Stored analyzed data (.pkl files)
│   └── Figures/                    # Auto-generated figures
│
├── environment.yml                 # Python environment specification
└── README.md                       # This file

• To analyze your own trajectories, modify the input paths in the notebook or analysis script to point to your topology and trajectory files.
• The workflow has been tested on Linux and should be portable to macOS or Windows if AMBER and Python dependencies are available.
• GPU acceleration (CUDA) is highly recommended for production-length simulations.

The solvent-stripped trajectories are deposited on Zenodo 10.5281/zenodo.17973165.

Citation:
Structures of ALG3, ALG9, and ALG12 reveal the assembly logic of the N-glycan oligomannose core. J. Andrew N. Alexander, Shu-Yu Chen, Somnath Mukherjee, Mario de Capitani, Rossitza N. Irobalieva, Lorenzo Rossi, Parth Agrawal, Julia Kowal, Matheus A. Meirelles, Markus Aebi, Jean-5 Louis Reymond, Anthony A. Kossiakoff, Sereina Riniker and Kaspar P. Locher (2025) Nature Biochemisrtry XXX.XXX*