Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM$^3$-DTI: A Large Language Model and Multi-modal data co-powered framework for Drug-Target Interaction prediction

Yuhao Zhang, Qinghong Guo, Qixian Chen, Liuwei Zhang, Hongyan Cui, Xiyi Chen

TL;DR

DTI prediction benefits from incorporating rich textual descriptions of drugs and targets alongside structural topology. The authors propose $LLM^3$-DTI, a framework that uses a domain-specific LLM to encode textual data and a dual cross-attention mechanism with TSFusion to align and fuse multi-modal embeddings for improved predictions. Experiments on a dataset with 708 drugs and 1,493 targets show state-of-the-art performance across ACC, AUROC, AUPR, MCC, and F1, with ablation studies confirming the value of each component and analyses on imbalanced data and cold-start scenarios. The work advances DTI prediction and enables more reliable drug repurposing, with code and data publicly available.

Abstract

Drug-target interaction (DTI) prediction is of great significance for drug discovery and drug repurposing. With the accumulation of a large volume of valuable data, data-driven methods have been increasingly harnessed to predict DTIs, reducing costs across various dimensions. Therefore, this paper proposes a $\textbf{L}$arge $\textbf{L}$anguage $\textbf{M}$odel and $\textbf{M}$ulti-$\textbf{M}$odel data co-powered $\textbf{D}$rug $\textbf{T}$arget $\textbf{I}$nteraction prediction framework, named LLM$^3$-DTI. LLM$^3$-DTI constructs multi-modal data embedding to enhance DTI prediction performance. In this framework, the text semantic embeddings of drugs and targets are encoded by a domain-specific LLM. To effectively align and fuse multi-modal embedding. We propose the dual cross-attention mechanism and the TSFusion module. Finally, these multi-modal data are utilized for the DTI task through an output network. The experimental results indicate that LLM$^3$-DTI can proficiently identify validated DTIs, surpassing the performance of the models employed for comparison across diverse scenarios. Consequently, LLM$^3$-DTI is adept at fulfilling the task of DTI prediction with excellence. The data and code are available at https://github.com/chaser-gua/LLM3DTI.

LLM$^3$-DTI: A Large Language Model and Multi-modal data co-powered framework for Drug-Target Interaction prediction

TL;DR

DTI prediction benefits from incorporating rich textual descriptions of drugs and targets alongside structural topology. The authors propose -DTI, a framework that uses a domain-specific LLM to encode textual data and a dual cross-attention mechanism with TSFusion to align and fuse multi-modal embeddings for improved predictions. Experiments on a dataset with 708 drugs and 1,493 targets show state-of-the-art performance across ACC, AUROC, AUPR, MCC, and F1, with ablation studies confirming the value of each component and analyses on imbalanced data and cold-start scenarios. The work advances DTI prediction and enables more reliable drug repurposing, with code and data publicly available.

Abstract

Drug-target interaction (DTI) prediction is of great significance for drug discovery and drug repurposing. With the accumulation of a large volume of valuable data, data-driven methods have been increasingly harnessed to predict DTIs, reducing costs across various dimensions. Therefore, this paper proposes a arge anguage odel and ulti-odel data co-powered rug arget nteraction prediction framework, named LLM-DTI. LLM-DTI constructs multi-modal data embedding to enhance DTI prediction performance. In this framework, the text semantic embeddings of drugs and targets are encoded by a domain-specific LLM. To effectively align and fuse multi-modal embedding. We propose the dual cross-attention mechanism and the TSFusion module. Finally, these multi-modal data are utilized for the DTI task through an output network. The experimental results indicate that LLM-DTI can proficiently identify validated DTIs, surpassing the performance of the models employed for comparison across diverse scenarios. Consequently, LLM-DTI is adept at fulfilling the task of DTI prediction with excellence. The data and code are available at https://github.com/chaser-gua/LLM3DTI.

Paper Structure

This paper contains 26 sections, 12 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Material and methods
  3. Multi-modal Embedding Construction
  4. Structural topology embedding
  5. Text semantic embedding
  6. Dual Cross-attention Alignment
  7. Text and Structure embedding Fusion
  8. Prediction block and loss function
  9. Experiment Results
  10. Datasets
  11. Baselines
  12. Evaluation metrics
  13. Experiment settings
  14. Analysis of performance
  15. Ablation study
  16. ...and 11 more sections

Figures (7)

  • Figure 1: The overall framework we proposed.
  • Figure 2: Ablation study results.
  • Figure 3: Parameter sensitivity analysis.
  • Figure 4: Imbalanced data training performance.
  • Figure 5: Cold start scenario performance.
  • ...and 2 more figures