[論文レビュー] Large Language Models in Drug Discovery and Development: From Disease Mechanisms to Clinical Trials

Yizhen Zheng, Huan Yee Koh+6 more|arXiv (Cornell University)|Sep 6, 2024

Computational Drug Discovery Methods被引用数 12

ひとこと要約

この論文は、LLMsが疾病機構理解、薬物発見、臨床試験の各段階でどのように統合され得るかを概観し、パラダイム、進展、今後の方向性を整理する。

ABSTRACT

The integration of Large Language Models (LLMs) into the drug discovery and development field marks a significant paradigm shift, offering novel methodologies for understanding disease mechanisms, facilitating drug discovery, and optimizing clinical trial processes. This review highlights the expanding role of LLMs in revolutionizing various stages of the drug development pipeline. We investigate how these advanced computational models can uncover target-disease linkage, interpret complex biomedical data, enhance drug molecule design, predict drug efficacy and safety profiles, and facilitate clinical trial processes. Our paper aims to provide a comprehensive overview for researchers and practitioners in computational biology, pharmacology, and AI4Science by offering insights into the potential transformative impact of LLMs on drug discovery and development.

研究の動機と目的

• Define the two main LLM paradigms (specialized vs general) in drug discovery and development.
• Outline a three-stage drug development pipeline (Understanding Disease Mechanisms、Drug Discovery、Clinical Trials) and map LLM capabilities to each stage.
• Assess current maturity of LLM applications across stages (not applicable, nascent, advanced, mature).
• Identify future directions and ethical, privacy, fairness, and bias considerations in LLM deployment.

提案手法

• Classify LLM types and categorize the drug development pipeline into three stages with corresponding tasks LLMs can perform.
• Evaluate applications of LLMs across stages and assign maturity levels (not applicable, nascent, advanced, mature) as illustrated in Figure 6.
• Synthesize existing literature on specialized nucleotide LLMs, transcriptomic LLMs, and protein-target analysis to discuss capabilities and limitations.
• Discuss technical challenges (hallucinations, context window, interpretability) and propose directions for trustworthy deployment.

実験結果

リサーチクエスチョン

• RQ1 How can LLMs be effectively integrated into each stage of drug discovery and development?
• RQ2 How advanced are LLMs in supporting downstream tasks across disease understanding, discovery, and clinical trials?
• RQ3 What are the future directions, challenges, and ethical considerations for LLMs in drug development?

主な発見

• LLMs exist in two main paradigms: specialized LLMs trained on scientific languages and general-purpose LLMs trained on broad text.
• LLMs can aid understanding of disease mechanisms through literature reviews, target-disease linkage analysis, and target validation.
• In drug discovery, specialized LLMs assist in chemistry tasks, ADMET prediction, retrosynthesis, and molecule generation/editing; general LLMs enable broader reasoning and workflow assistance.
• In clinical trials, LLMs can support patient-trial matching, trial planning, outcome prediction, and document generation.
• Genomics and transcriptomics applications include nucleotide LLMs (e.g., DNA-BERT), gene network analysis (Geneformer), and sparse-data adaptations for transcriptomics.
• Protein-target analysis leverages LLMs for evolutionary conservation, protein folding insights, binding site prediction, and structure prediction through models like ESM, AlphaFold2, and RosettaFold.