DeepCancerMap

DeepCancerMap predicts inhibitory activity of small molecules against anticancer targets and cancer cell lines using deep learning to support target- and cell-based anticancer drug discovery.

Key Features:

• Extensive Data Integration: Integrates 485,900 compounds and 3,919,974 bioactivity records against 426 anticancer targets and 346 cancer cell lines from academic literature, and includes data from 60 tumor cell lines of the NCI-60 panel.
• Advanced Predictive Modeling: Implements FP-GNN (Fragment-based Graph Neural Network) to construct 832 classification models, comprising 426 target-based and 406 cell-based models that predict inhibitory activity.
• Superior Predictive Performance: FP-GNN models report Area Under the Curve (AUC) values of 0.91 for target-based predictions, 0.88 for academia-sourced cancer cell line predictions, and 0.91 for NCI-60 cancer cell line predictions, outperforming traditional machine learning and other deep learning approaches.

Scientific Applications:

• Target-based screening: Predicts compound inhibition against specific anticancer targets to support target-based screening campaigns.
• Phenotypic/cell-based screening: Predicts compound activity across cancer cell lines to support phenotypic and cell-based screening.
• Large-scale virtual screening: Enables large-scale virtual screening of compound libraries for anticancer activity prediction.
• Target fishing: Supports target fishing by predicting probable target interactions of compounds.
• Drug repositioning: Facilitates drug repositioning by profiling compounds across targets and cell lines to identify potential new indications.

Methodology:

FP-GNN (Fragment-based Graph Neural Network) deep learning was used to construct 832 classification models trained on an integrated dataset of 485,900 compounds and 3,919,974 bioactivity records covering 426 targets, 346 academic-sourced cell lines, and 60 NCI-60 cell lines.