DFT_ANPD :A Dual-Feature Two-sided Attention Network for Anticancer Natural Products Detection

Objectives

Cancer remains the leading cause of death globally despite progress in research. Natural products have historically played a key role in anticancer drug discovery. Experimental and computational methods are used to identify anticancer compounds. This study introduces DFT_ANPD, a deep learning framework for predicting anticancer properties in natural compounds. It integrates molecular structural data with embeddings from large language models (LLMs). A 1D-CNN processes SMILES strings for chemical patterns, while a fine-tuned SMILES-BERT extracts semantic information. These insights are fused using a two-sided attention mechanism to prioritize relevant features, enabling prediction via a sigmoid activation function.

Findings

DFT_ANPD outperformed existing methods, like iANP-EC, on NPACT, CancerHSP, and NPASS datasets, showing superior accuracy and generalization.Actinomycin D was identified, proving the model's robustness and clinical relevance.

Installation Guide

1. Prerequisites

Ensure you have the following installed on your system:

• Python 3.7 or newer
• pandas
• numpy
• rdkit
• scikit-learn
• tensorflow
• keras
• torch
• matplotlib
• seaborn
• pyswarm

2. Clone the Repository

git clone https://github.com/Rambono/DFT_ANPD.git

cd DFT_ANPD

Components

• Generalization_data preprocess.py - Preprocess Generalization Experiment
• Comparison.py - Compare our model with previous model
• First experiment - First Experiment in Generalization
• second experiment - Second Experiment in Generalization
• experiment_one - DFT_ANPD with 5 Fold Cross validation
• experiment_two - DFT_ANPD with independent test set
• confusion_matrix_plot.py - confusion_matrix plotting
• Evaluation metrics plot.py - Evaluation metrics plotting
• common and unique smiles - compare NPASS with NPACT and Cancer HSP data