https://github.com/maryamastero/moleculepropertyprediction

This repository aims to leverage pretrained neural networks for predicting node labels in molecular graphs, with a focus on fine-tuning for molecular property prediction tasks. The dataset used for this purpose is Zinc.
https://github.com/maryamastero/moleculepropertyprediction

deep-learning gnn molecule

Last synced: 22 days ago
JSON representation

This repository aims to leverage pretrained neural networks for predicting node labels in molecular graphs, with a focus on fine-tuning for molecular property prediction tasks. The dataset used for this purpose is Zinc.

• Host: GitHub
• URL: https://github.com/maryamastero/moleculepropertyprediction
• Owner: maryamastero
• Created: 2024-06-27T14:37:54.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-06-27T14:54:16.000Z (7 months ago)
• Last Synced: 2024-11-01T08:42:12.001Z (2 months ago)
• Topics: deep-learning, gnn, molecule
• Language: Python
• Homepage:
• Size: 30.3 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Comparison of Deep Learning Models for Molecular Property Prediction Using the ZINC Dataset

## 1. Objective

The primary goal of this work is to compare different deep learning models for molecular property prediction using the ZINC dataset.

## 2. Methodology

**Dataset**: The ZINC dataset is a collection of commercially available chemical compounds. It provides molecular structures and associated properties, which are used for training and evaluating the models.

**Model Descriptions**:

- **Model 1: GIN (Graph Isomorphism Network)**:

  - A GIN is designed to predict molecular properties directly from the graph representation of molecules.

- **Model 2: Pretrained and Fine-tuned Model**:

  - This model is first pretrained on a node classification task to learn general features of molecular graphs.

  - It is then fine-tuned specifically for the property prediction task.

- **Model 3: Multitask Learning Model**:

  - This model simultaneously learns two tasks: node classification and property prediction.

  - It shares some components between the tasks to leverage multitask learning benefits.

**Code Attribution**: This code is adapted from [deepfindr/gnn-project](https://github.com/deepfindr/gnn-project).

## 3. Experimental Setup

**Training and Validation**:

- The training procedure includes splitting the data into training, validation, and test sets.

- Hyperparameters, loss functions, and optimization techniques are kept consistent across models for fair comparison.

**Evaluation Metrics**:

- The primary evaluation metric is Mean Squared Error (MSE) for property prediction.

## 4. Results

**Performance Comparison**:

- The following table summarizes the MSE results for all three models:

| Model                                | Mean Squared Error (MSE) |

|--------------------------------------|--------------------------|

| GIN                                  | 0.133                    |

| Pretrained and Fine-tuned            | 0.120                    |

| Multitask Learning                   | 0.143                    |

- The pretrained and fine-tuned model achieved the lowest MSE.