The properties of known drugs. 1. Molecular frameworks.

By G. Bemis, Mark A. Murcko. 1996

Subfields: Computational Chemistry & Drug Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Chemistry & Drug Discovery

Citations in the corpus

AtomNet: A Deep Convolutional Neural Network for Bioactivity Prediction in Structure-based Drug Discovery (2015, methodology): “Third, we partitioned the data into training, validation, and testing sets by ﬁrst clustering the active ligands for each target based on their Bemis-Murcko scaffolds [28] and choosing ligands that were at least 3 µM apart as the cluster exemplars.”
Multi-Scale Representation Learning on Proteins (2022, methodology): “The scaffold split is prepared by computing the Bemis-Murcko scaffold (Bemis and Murcko, 1996) using RDKit for each molecule, and splitting the molecules such that molecules with rare or unseen scaffolds are part”
Unsupervised learning of dynamical and molecular similarity using variance minimization (2017, background): “Common partitions for chemoinformatic studies include random splits, temporal splits, stratiﬁed splits according to the quantity being learned, and scaffold splits [Bemis and Murcko, 1996] in which molecules are assigned to a set based on the frequency of the molecular scaffold.”
Molecular Shape and Medicinal Chemistry: A Perspective (2010, background): “88,89 In fact, as shown in Figure 17, 500 of the most popular reference shapes cover about 90% of the 3 million molecules.”
A strategy for structuring and reporting a read-across prediction of toxicity. (2015, background): “…into groups of atoms representing rings (e.g., benzene or naphthalene), linkers (i.e., atoms in a direct path connecting two ring systems), frameworks (i.e., the combination of ring systems and linkers in a molecule), and side chains (i.e., non-ring, non-linker atoms) (Bemis and Murcko, 1996).”
Geometry-enhanced molecular representation learning for property prediction (2021, methodology): “There are several popular splitting methods used in molecular property prediction datasets, including 394 random splitting, scaffold splitting [2] and random scaffold splitting.”
Pre-training Graph Neural Networks (2019, methodology): “For molecular prediction tasks, following [38], we cluster molecules by graph substructure [4], and recombine the clusters by placing the most common scaffolds (molecular substructures) in the training set, producing validation and test sets that contain structurally different molecules.”