PfHT1

Substrate binding reorganizes the energetic landscape of Plasmodium falciparum hexose transporter PfHT1

---

Overview

This repository accompanies our PfHT1 molecular mechanism study and contains:

• Graph Attention Network (GAT) model
• Transition-path theory (TPT), pathway fingerprinting, and VAE model scripts
• custom MSM, free-energy, residue-contact, binding-pocket/TM7b, and yeast-growth analyses
• preprocessing scripts for graph and transition-path generation
• pretrained model weights

Large datasets including molecular dynamics trajectories, topology files, MSM data, and processed simulation datasets are hosted separately in the associated data repository.

---

---

Repository Contents

PfHT1/
├── preprocessing/
│   ├── make_input_graphs.py
│   └── tpt_fingerprints.py
├── GAT_Model/
├── VAE_Model/
│   ├── analysis_config.example.json
│   ├── train_vae.py
│   └── cluster_pathways.py
├── custom_analysis_codes/
│   ├── msm_kinetics/
│   ├── free_energy_landscapes/
│   ├── binding_pocket_tm7b/
│   ├── residue_contacts/
│   └── yeast_growth/
├── requirements.txt
└── figures/

---

Workflow

Generate graph inputs:

export PFHT1_DATA_DIR=/path/to/shared_data

python preprocessing/make_input_graphs.py \
    --trajectory "$PFHT1_DATA_DIR/GAT/filtered_trajectory.xtc" \
    --topology "$PFHT1_DATA_DIR/topology/PfHT1_OF_with_glc_mix_memb-strip.prmtop" \
    --output "$PFHT1_DATA_DIR/GAT/processed_graphs/ligand_residue_allosteric_graphs_proteinCOM.pt"

The notebook GAT_Model/gat_interpretable-residue_importance.ipynb reads the same data location from PFHT1_DATA_DIR and contains:

• model architecture
• training workflow
• inference and evaluation

Pretrained weights:

GAT_Model/best_model.pt

TPT--VAE pathway analysis

Copy VAE_Model/analysis_config.example.json, supply the apo and holo MSM, discrete-trajectory, feature, and tICA input paths, and run:

python preprocessing/tpt_fingerprints.py \
    --config VAE_Model/analysis_config.json

python VAE_Model/train_vae.py \
    --system apo --seed 167 \
    --fingerprints outputs/fingerprints/apo.npz \
    --output outputs/models/apo

python VAE_Model/train_vae.py \
    --system holo --seed 239 \
    --fingerprints outputs/fingerprints/holo.npz \
    --output outputs/models/holo

python VAE_Model/cluster_pathways.py \
    --embedding outputs/models/apo/embedding.npz \
    --fingerprints outputs/fingerprints/apo.npz \
    --output outputs/clusters/apo

Repeat the clustering command with the holo files. Generated TPT paths, fingerprints, trained models, and clustering outputs are not tracked in Git.

Custom molecular-simulation and experimental analyses

The custom_analysis_codes/ directory contains the remaining custom analysis used in the manuscript. Apo and glucose-bound calculations share the same configurable scripts instead of maintaining condition-specific copies. The included workflows cover:

• transport-coordinate featurization, tICA, MSM construction, bootstrapping, and macrostate MFPTs
• MSM-weighted free-energy landscapes
• fresh MSM-weighted sampling and apo-reference clustering of binding-pocket geometries
• TM7b hydrogen-bond cracking and the corresponding two-state equilibrium free-energy difference
• residue-residue contact-score differences
• pooled yeast-growth analysis

See custom_analysis_codes/README.md for inputs, analysis definitions, parameter values, and execution order.

---

Additional Data

The following files are hosted separately due to size limitations:

• filtered glucose-transport trajectory used for GAT graph preprocessing
• topology (.prmtop) files
• MSM datasets
• processed graph datasets
• framewise feature and tICA datasets used by the TPT--VAE analysis
• trajectories, MSM models, frame-provenance tables, and intermediate arrays used by the custom analyses

Data repository: (https://uofi.box.com/s/i5l1juaejny6m4bet8ybkmpyp5xqrdfs)

---

Citation

If you use this repository, please cite: Paul, A., Xu, L., & Shukla, D. (2026). Substrate binding reorganizes the energetic landscape of Plasmodium falciparum hexose transporter PfHT1. bioRxiv, 2026-09

---