分子指紋による薬物相互作用の予測におけるモデルの過度に複雑さを解決する
Manel Gil-Sorribes1, Alexis Molina2
1Nostrum Biodiscovery, Barcelona, 08029, Spain.
Journal of cheminformatics
|February 19, 2026
まとめ
薬物相互作用を予測することは極めて重要です. モルガン指紋 (MFP) のような単純な分子表現は,多くのベンチマークで複雑なモデルに匹敵し,または超え,より良いデータセットと評価が薬物安全性研究の鍵であることを示唆しています.
科学分野:
- 計算化学はコンピュータ化学である.
- 薬理学 薬理学とは
- 薬剤発見における機械学習
背景:
- 薬物相互作用 (DDI) の正確な予測は,医薬品研究と患者の安全性にとって極めて重要です.
- 現在のトレンドは複雑なモデルを好むが,標準ベンチマークのリターンは低下している.
- 分子表現の評価は,モデルのパフォーマンスを理解するための鍵です.
研究 の 目的:
- DDI予測の精度に対する分子表現の影響を隔離する.
- モルガン指紋 (MFP),グラフコンボリューションネットワーク (GCN),およびMoLFormerの埋め込みのパフォーマンスを比較する.
- データセットの分割がモデル評価に及ぼす影響を調査する.
主な方法:
- 固定された分類器アーキテクチャが使用され,分子表現のみが交換されました.
- ECFP4 モルガン指紋 (MFP),事前訓練されたGCN,およびMoLFormerの埋め込みを比較しました.
- DrugBankのDDI分割とFDAの薬物関連性ベンチマークのモデルを評価し,漏れ防止と流通外分割を含む.
主要な成果:
- 浅いヘッドを持つMFPは,より少ないパラメータを使用して,標準のDrugBankとFDAのベンチマークでより複雑なモデルに匹敵し,またはそれを上回りました.
- Unseen DDIの分割では,MFPはAUROC 99.4とAUPR 98.4を達成し,MoLFormerと以前の最先端を上回った.
- 予備訓練されたGCNは,配送外での厳格な支架分割 (AUROC 73.99) をリードし,挑戦的なシナリオにおけるモデル能力の利点を強調しました.
結論:
- MFPのような単純な分子表現は,DDI予測に非常に競争力があります.
- モデルの複雑性の増加による絶対的利益は,標準的ベンチマークでは最小限です.
- DDIの予測における将来の進歩は,より複雑なモデルではなく,データセットの改善されたキュレーションと厳格なアウトディストリビューション評価に依存しています.
関連する概念動画
Protein-protein Interfaces
14.8K
Many proteins form complexes to carry out their functions, making protein-protein interactions (PPIs) essential for an organism's survival. Most PPIs are stabilized by numerous weak noncovalent chemical forces. The physical shape of the interfaces determines the way two proteins interact. Many globular proteins have closely-matching shapes on their surfaces, which form a large number of weak bonds. Additionally, many PPIs occur between two helices or between a surface cleft and a...
14.8K
Quantitative Aspects of Drug-Receptor Interaction
1.9K
The receptor occupancy theory connects a drug's response to the number of occupied receptors. With higher drug concentrations, more receptors are occupied, leading to increased responses. The formation of drug-receptor complexes involves association and dissociation rates, which reach equilibrium when the forward and backward reactions are equal. The equilibrium association constant (Ka) and its inverse, the equilibrium dissociation constant (Kd), indicate drug affinity. Higher Ka and lower...
1.9K
Drug Discovery: Overview
12.1K
Drug discovery is a multifaceted process involving extensive screening, testing, and optimization of lead compounds to identify potential new drugs for therapeutic use. It combines several approaches, including screening large numbers of natural products, chemical modification of known active molecules, identification of new drug targets, and rational design based on biological mechanisms and drug-receptor structure. These approaches are carried out in both academic research laboratories and...
12.1K
Pharmacodynamic Models: Additive and Proportional Drug Effect Model
23
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...
23
Pharmacokinetic Models: Overview
2.2K
Pharmacokinetic models utilize mathematical analysis to achieve a detailed quantitative understanding of a drug's life cycle within the body. They are instrumental in simulating a drug's pharmacokinetic parameters, predicting drug concentrations over time, optimizing dosage regimens, linking concentrations with pharmacologic activity, and estimating potential toxicity.
There are three primary types of models: empirical, compartment, and physiological. Empirical models, with minimal...
There are three primary types of models: empirical, compartment, and physiological. Empirical models, with minimal...
2.2K
Pharmacodynamic Models: Link Model and Systems Pharmacodynamic Model
25
The link model is a fundamental pharmacokinetic-pharmacodynamic (PK–PD) approach to account for delayed drug responses when the observed effect does not immediately correlate with the drug's plasma concentration peak. This delay is mathematically addressed by introducing an effect compartment concentration, Ce, which is kinetically linked to the plasma concentration, Cp, via a first-order rate constant, ke0. The linkage allows for a more accurate prediction of drug effects over time. A...
25


