De Novo Antioxidant Peptide Design Via Machine Learning and DFT studies

About

Welcome to the repository for our article: De Novo Antioxidant Peptide Design Via Machine Learning and DFT Studies. This project revolves around the development of a deep generative model, leveraging GRU layers, to create antioxidant peptides.

Project Overview

1. Pretrained Generative Model: We kick-started the project by crafting a pretrained generative model using TensorFlow. Refer to 02_GRU_Base.ipynb for detailed insights.

2. Fine-Tuning for Antioxidant Peptides: We then fine-tuned this model to tailor its focus specifically towards generating antioxidant peptides. The fine-tuning process is documented in 03_GRU_TL.ipynb.

3. Peptide Generation and Classification: Utilizing the fine-tuned model, we generated new peptide sequences (refer to 04_Generate_data_TL.ipynb). Moreover, to predict antioxidant activity, we developed a classification model outlined in 05_Conv1d_Classification.ipynb.

4. Filtering and Synthesis: Following generation and classification, we fiterd the generated sequences (06_filter_gen_data.ipynb, 07_analysis_filter_cluster.ipynb) based on various criteria. The remaining peptides were then synthesized for further activity assessment.

Dataset Availability

• The base model training dataset can be found here.
• For training the fine-tuned generative and classification models, we utilized the dataset available here.
• We also used peptipedia's dataset to check our model's output novelty here