CaliciBoost:对 caco-2 透性预测分子表示的性能驱动评估
Huong Van Le1, Weibin Ren1, Junhong Kim1
1Calici Co., Ltd, Korea, Daejeon, 34138, Republic of Korea.
Journal of cheminformatics
|December 23, 2025
概括
使用分子描述符的自动机器学习显著改善了用于药物发现的Caco-2透性预测. 整合3D描述符可以提高准确性,从而在公共排行榜上获得最佳表现.
科学领域:
- 计算化学和化学信息学
- 药物的发现和开发.
- 机器学习在毒理学中的应用
背景情况:
- 可可-2 透性是预测口服药物吸收的关键体外试验.
- 准确的计算预测Caco-2的透性对于有效的早期药物发现至关重要.
- 现有的方法需要对各种分子表示和机器学习技术进行系统的评估.
研究的目的:
- 系统地对八种分子特征表示类型进行基准测试,用于Caco-2的透性预测.
- 通过使用自动机器学习 (AutoML) 评估2D/3D描述符,指纹和深度学习嵌入的影响.
- 为了确定数据有限的预测任务的最佳特征选择策略.
主要方法:
- 研究了八种分子特征表示类型 (2D/3D描述符,指纹,嵌入).
- 采用自动机器学习 (AutoML) 技术进行模型训练和优化.
- 在两个数据集上评估模型:TDC基准和精选的OCHEM数据.
- 使用特征重要性分析 (SHAP,排列重要性) 进行解释性.
主要成果:
- 帕德尔,莫德雷德和RDKit描述符对Caco-2透性预测非常有效.
- 开发的CalciBoost模型实现了最低的平均绝对误差 (MAE),并在TDC Caco-2排行榜上排名第一.
- 整合3D描述符通常会比单独的2D特征提高预测性能.
- 功能重要性分析提供了对描述器贡献的见解.
结论:
- 自动ML方法是有效的预测ADMET属性,如Caco-2的透性.
- 3D分子描述器在提高预测准确性方面发挥着至关重要的作用.
- 该研究为化学信息学中的特征选择提供了实际指导,特别是在数据有限的场景中.
- 开发了一个基于PADEL的强大,可解释的AutoML模型,实现了最先进的性能.
相关概念视频
01:44The Significance of Membrane Transport
40.9K
The transport of solutes across the cell membrane is essential for metabolic processes, like maintaining cell size and volume, generating the action potential, exchanging nutrients and gases, etc. Membrane transport can be either passive or active. It can be simple diffusion, facilitated, or mediated transport aided by transport proteins such as transporters and channels.
Transporters facilitate either an active or passive movement of solutes. They can allow a single-molecule transport down its...
Transporters facilitate either an active or passive movement of solutes. They can allow a single-molecule transport down its...
40.9K
01:24Cellular Membranes and Drug Transport
1.3K
Drugs must traverse multiple biological barriers, such as multi-layered skin, single-layered intestinal epithelium, and the plasma membrane, to reach their target sites within the body. The plasma membrane, a highly structured composite of phospholipids, carbohydrates, and proteins, is the cell's protective boundary, facilitating selective substance exchange.
Phospholipids arrange themselves into a bilayer, with hydrophilic heads oriented outward and hydrophobic tails facing inward.
Phospholipids arrange themselves into a bilayer, with hydrophilic heads oriented outward and hydrophobic tails facing inward.
1.3K
01:23Drug Absorption Mechanism: Passive Membrane Transport
6.2K
Passive transport is a method of drug absorption where small, lipid-soluble drugs can move across the cell membrane. This movement happens along the concentration gradient, which is a natural flow from higher to lower concentration areas. The speed at which the drug moves is directly related to its lipid–water partition coefficient. This means that the more a drug dissolves in lipids, the faster it diffuses or spreads throughout the body. It is important to note that most drugs are either...
6.2K
01:20Factors Affecting Dissolution: Drug Permeability, Stability and Stereochemistry
477
Orally administered drugs primarily enter the systemic circulation via passive diffusion through the intestinal membranes. The drug's absorption is influenced by drug stability in the gastrointestinal GI tract, membrane permeability, the surface area available for absorption, luminal drug concentration, and residence time in the lumen. Drug permeability can be enhanced by adjusting the lipophilicity, polarity, or molecular size of the drug, promoting its passive transport across intestinal...
477
01:23Membrane Fluidity
171.9K
Cell membranes are composed of phospholipids, proteins, and carbohydrates loosely attached to one another through chemical interactions. Molecules are generally able to move about in the plane of the membrane, giving the membrane its flexible nature called fluidity. Two other features of the membrane contribute to membrane fluidity: the chemical structure of the phospholipids and the presence of cholesterol in the membrane.
171.9K
01:26Membrane Fluidity
14.4K
Membrane fluidity is explained by the fluid mosaic model of the cell membrane, which describes the plasma membrane structure as a mosaic of components—including phospholipids, cholesterol, proteins, and carbohydrates—that gives the membrane a fluid character.
Mosaic nature of the membrane
The mosaic characteristic of the membrane helps the plasma membrane remain fluid. The integral proteins and lipids exist as separate but loosely-attached molecules in the membrane. The membrane is...
Mosaic nature of the membrane
The mosaic characteristic of the membrane helps the plasma membrane remain fluid. The integral proteins and lipids exist as separate but loosely-attached molecules in the membrane. The membrane is...
14.4K