提高CYP3A4抑制预测使用混合GNN-ML模型与数据增强
Somin Woo1,2,3, Ju-Hyeok Jeon4, Sangil Han5
1Graduate School of Pharmacy, Kyungpook National University, Daegu 41566, Republic of Korea.
Pharmaceuticals (Basel, Switzerland)
|February 27, 2026
概括
这项研究开发了一个AI框架来预测细胞染色体P450 3A4 (CYP3A4) 抑制,这对于评估药物相互作用风险至关重要. 结合机器学习和图形神经网络的综合方法在代谢查中显示出更好的准确性和可解释性.
科学领域:
- 计算化学计算化学
- 药理学 药理学是指药理学的学科.
- 人工智能在药物发现中的作用
背景情况:
- 细胞染色体P450 3A4 (CYP3A4) 是药物代谢中的关键酶,影响30-50%的市场药物.
- 准确预测CYP3A4抑制对于早期识别潜在的药物相互作用 (DDI) 风险和毒性至关重要.
- 现有的预测方法需要加强,以提高药物开发管道的准确性和可解释性.
研究的目的:
- 评估一个集成的人工智能 (AI) 框架,用于预测CYP3A4抑制百分比.
- 评估各种机器学习 (ML) 和图形神经网络 (GNN) 模型的性能.
- 为代谢查和DDI风险评估建立一个可靠和可解释的框架.
主要方法:
- 从各种化学和公共数据库中编制了23713种化合物的大型数据集.
- 评估了基于矢量的ML模型 (LightGBM,XGBoost,CatBoost,加权组合) 和GNN模型 (O-GNN+CL+Mixup,D-MPNN,GINE,GATv2).这些模型是基于矢量的ML模型.
- 采用GNN培训的多元组混合和基于计数的SMILES测试时间增长用于推理,通过加权集团集成最佳模型.
主要成果:
- 权重的ML组合在ML模型中实现了最高的性能 (RMSE=19.1031,PCC=0.7566).
- 在O-GNN + CL + Mixup模型中,GNN中表现最好 (RMSE=20.1002,PCC=0.7265).
- 综合混合模型在外部验证后显示出卓越的预测准确性 (RMSE=19.0784,PCC=0.7570) 和可概括性.
结论:
- 将ML和GNN模型与数据增强策略集成显著提高了CYP3A4抑制预测的稳定性和可解释性.
- 开发的AI框架为早期代谢查提供了实用和有效的工具.
- 这种方法有助于主动评估DDI风险,为更安全的药物开发做出贡献.
相关概念视频
Model Approaches for Pharmacokinetic Data: Distributed Parameter Models
299
Pharmacokinetic models are mathematical constructs that represent and predict the time course of drug concentrations in the body, providing meaningful pharmacokinetic parameters. These models are categorized into compartment, physiological, and distributed parameter models.
The distributed parameter models are specifically designed to account for variations and differences in some drug classes. This model is particularly useful for assessing regional concentrations of anticancer or...
The distributed parameter models are specifically designed to account for variations and differences in some drug classes. This model is particularly useful for assessing regional concentrations of anticancer or...
299
Pharmacodynamic Models: Additive and Proportional Drug Effect Model
41
Drug response models describe how pharmacological agents interact with biological systems to produce measurable effects. Baseline responses are inherent physiological activities without a drug significantly influencing the observed pharmacological outcomes. Depending on the drug response model employed, these baseline responses may combine with the drug's effect in either an additive or proportional manner.Additive Drug Response ModelIn the additive model, the drug effect is independent of the...
41
Model Approaches for Pharmacokinetic Data: Compartment Models
653
Compartmental analysis is a widely adopted approach to characterizing drug pharmacokinetics. It uses compartment models that conceptualize the body as a collection of reversibly communicating compartments, each representing a group of tissues exhibiting similar drug distribution characteristics. The movement rate of the drug between these compartments is typically described by first-order kinetics.
Two primary types of compartment models are recognized: mammillary and catenary. The more...
Two primary types of compartment models are recognized: mammillary and catenary. The more...
653
