Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Combining GCN Structural Learning with LLM Chemical Knowledge for Enhanced Virtual Screening

Radia Berreziga, Mohammed Brahimi, Khairedine Kraim, Hamid Azzoune

TL;DR

This work tackles the limitations of descriptor-based and graph-only approaches in virtual screening by introducing a GCN-LLM hybrid that injects precomputed SMILES-derived embeddings into every GCN layer. The model leverages a precomputed molecular feature library to maintain efficiency while jointly learning structural and semantic information, achieving superior performance across six kinase-target datasets. The key contributions include the architectural design of a six-layer GCN with layer-wise SMILES fusion, a comprehensive dataset preprocessing and labeling pipeline, and demonstrated improvements in accuracy, F1-score, and AUC-ROC. The approach advances multimodal molecular representations and offers a scalable path toward more accurate and explainable virtual screening with practical implications for drug discovery.

Abstract

Virtual screening plays a critical role in modern drug discovery by enabling the identification of promising candidate molecules for experimental validation. Traditional machine learning methods such, as Support Vector Machines (SVM) and XGBoost, rely on predefined molecular representations, often leading to information loss and potential bias. In contrast, deep learning approaches-particularly Graph Convolutional Networks (GCNs)-offer a more expressive and unbiased alternative by operating directly on molecular graphs. Meanwhile, Large Language Models (LLMs) have recently demonstrated state-of-the-art performance in drug design, thanks to their capacity to capture complex chemical patterns from large-scale data via attention mechanisms. In this paper, we propose a hybrid architecture that integrates GCNs with LLM-derived embeddings to combine localized structural learning with global chemical knowledge. The LLM embeddings can be precomputed and stored in a molecular feature library, removing the need to rerun the LLM during training or inference and thus maintaining computational efficiency. We found that concatenating the LLM embeddings after each GCN layer-rather than only at the final layer-significantly improves performance, enabling deeper integration of global context throughout the network. The resulting model achieves superior results, with an F1-score of (88.8\%), outperforming standalone GCN (87.9%), XGBoost (85.5%), and SVM (85.4%) baselines.

Combining GCN Structural Learning with LLM Chemical Knowledge for Enhanced Virtual Screening

TL;DR

This work tackles the limitations of descriptor-based and graph-only approaches in virtual screening by introducing a GCN-LLM hybrid that injects precomputed SMILES-derived embeddings into every GCN layer. The model leverages a precomputed molecular feature library to maintain efficiency while jointly learning structural and semantic information, achieving superior performance across six kinase-target datasets. The key contributions include the architectural design of a six-layer GCN with layer-wise SMILES fusion, a comprehensive dataset preprocessing and labeling pipeline, and demonstrated improvements in accuracy, F1-score, and AUC-ROC. The approach advances multimodal molecular representations and offers a scalable path toward more accurate and explainable virtual screening with practical implications for drug discovery.

Abstract

Virtual screening plays a critical role in modern drug discovery by enabling the identification of promising candidate molecules for experimental validation. Traditional machine learning methods such, as Support Vector Machines (SVM) and XGBoost, rely on predefined molecular representations, often leading to information loss and potential bias. In contrast, deep learning approaches-particularly Graph Convolutional Networks (GCNs)-offer a more expressive and unbiased alternative by operating directly on molecular graphs. Meanwhile, Large Language Models (LLMs) have recently demonstrated state-of-the-art performance in drug design, thanks to their capacity to capture complex chemical patterns from large-scale data via attention mechanisms. In this paper, we propose a hybrid architecture that integrates GCNs with LLM-derived embeddings to combine localized structural learning with global chemical knowledge. The LLM embeddings can be precomputed and stored in a molecular feature library, removing the need to rerun the LLM during training or inference and thus maintaining computational efficiency. We found that concatenating the LLM embeddings after each GCN layer-rather than only at the final layer-significantly improves performance, enabling deeper integration of global context throughout the network. The resulting model achieves superior results, with an F1-score of (88.8\%), outperforming standalone GCN (87.9%), XGBoost (85.5%), and SVM (85.4%) baselines.

Paper Structure

This paper contains 26 sections, 3 equations, 4 figures, 5 tables.

Table of Contents

  1. Abstract
  2. Introduction
  3. Related work
  4. Descriptor-based methods
  5. Graph representation approaches
  6. SMILES approaches
  7. Methodology
  8. GCN-LLM Model Architecture
  9. Datasets
  10. Enzyme Target and Dataset Source
  11. Data Preprocessing
  12. Labeling and Activity Classification
  13. Molecular Descriptor Generation and Feature Selection
  14. Dataset Summary
  15. Training and Evaluation
  16. ...and 11 more sections

Figures (4)

  • Figure 1: Architecture of the GCN-LLM model integrating SMILES encodings with graph features for molecular classification.
  • Figure 2: Bar plot comparing the accurancy of SVM, XGBoost, GCN, and GCN-LLM across six datasets.
  • Figure 3: Bar plot comparing the F1 Score of SVM, XGBoost, GCN, and GCN-LLM across six datasets.
  • Figure :