Application of Large Language Models in Drug-Induced Osteotoxicity Prediction

Area of Science:

• Pharmacology and Toxicology
• Computational Chemistry
• Drug Safety

Background:

• Drug-induced osteotoxicity presents significant clinical and developmental challenges.
• Current toxicity assessment models lack specialized datasets and algorithms for predicting bone toxicity.
• Identifying and mitigating skeletal side effects is crucial for patient safety.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting drug-induced osteotoxicity.
• To assess the efficacy of large language models (LLMs) in identifying osteotoxic molecules.
• To improve safety protocols in drug development through accurate toxicity prediction.

Main Methods:

• Collection of a dataset of osteotoxic molecules.
• Application of various machine learning algorithms, including DeepSeek and ChatGPT.
• Comparative analysis of LLMs and traditional machine learning methods for toxicity prediction.

Main Results:

• DeepSeek R1 and ChatGPT o3 models achieved high accuracy (ACC values of 0.87 and 0.88, respectively).
• Machine learning approaches demonstrated significant potential in predicting molecular osteotoxicity.
• LLMs showed promise in identifying harmful effects on bone health.

Conclusions:

• Machine learning, particularly LLMs, can effectively predict drug-induced osteotoxicity.
• These models offer valuable tools for enhancing drug safety evaluations during development.
• The findings support improved skeletal health monitoring and public safety in chemical and health sciences.