Code for [DeepPH : A Multimodal Deep Learning Model for Predicting Enzyme Optimal pH Range]

This supplementary code supports the results in our paper.

File Descriptions

• data_split.py
  Script for preprocess ing enzyme data into graph format with multimodal features for pH range prediction.

• egnn_clean.py
  Defines the core EGNN architecture used in our method.

• egnn_model_split_range.py
  The full model pipeline for training on range prediction tasks.

• train_model_value_split_range.py
  Main training script for learning both the pH value and its interval (range).
  Supports GPU training and logs intermediate metrics.

• test_model_r2_split_aa_range3.pt
  The trained model containing learned weights.

-‘new_train_value.pkl, new_test_value.pkl and new_test_value_remove_phenv.pkl' They are our training dataset and two test sets.

How to Run

1. Environment Setup
   The proposed model was implemented using Python 3.8+, PyTorch 2.6.0 and PyTorch Geometric 2.6.1.
2. Execute the following command to train the DeepPH model: 'python train_model_value_split_range.py'