Skip to main content
[Preprint]. 2023 Oct 18:2023.10.15.562398. [Version 1] doi: 10.1101/2023.10.15.562398

Table 2:

The description of various feature spaces used in this study.

Feature Space Dimensions after feature selection (where applicable) Description Signal Expected Source
Chemical Structure 2048-bit vector ECFP4 (Morgan) fingerprints representing chemical structures Distinctive patterns of chemical bonding and arrangement 52
Physicochemical properties (Mordred Descriptors) 1038 2-D descriptors Properties such as lipophilicity, solubility, molecular weight, ionizing potential, etc. Properties that are associated with negative impacts on ion channels in the heart 41
MOA dataset 264 binary encoded MOAs + 551 known targets Annotations for mechanism of action and known targets based on knowledge. Mechanism of action for drugs that inhibit certain ion channels 29
CellScape Target Prediction dataset 1893 predictions for 817 unique targets and concentration combinations (0.1, 1, 10, and 100uM) Predicted protein target for inhibition/antagonism; does not consider the functionality; prediction is based on chemical structure; updated algorithm from PIDGINv443 Understanding how a drug interacts with various biological targets (not just its primary target) can provide insights into potential off-target effects 34,43
Cell Painting 1783 features Morphological changes in U2OS cells by a chemical perturbation, using a 5-channel fluorescence microscopy assay Morphological changes in cells that reflect basic biological processes 24,53
Gene Expression 978 features Transcriptomic changes in response to chemicals using the L1000 assay Upregulation or downregulation of genes associated with cardiac stress, apoptosis in cardiac cells, or ion channel function 19,25
Gene Ontology 4438 annotations Gene Ontology manual annotations based on collective knowledge Understanding the biological processes, cellular components, and molecular functions affected, e.g., related to cardiac function, cardiac muscle tissue development, or ion homeostasis 19,26
Cmax 2 features The maximum total and unbound concentration of a drug in plasma High Cmax would indicate a high risk of cardiotoxicity 30