PaccMann-MCA
Prediction of AntiCancer Compound sensitivity with Multi-modal Attention-based Neural Networks
Model Architecture
The PaccMann-MCA model is composed of a network propagation layer, an attention-based gene expression encoder, a SMILES embedding layer and a SMILES encoder. The gene expression encoder generates attention weights to encode the cell-line gene expression. The model uses multiscale convolutional attention (MCA) mechanism for the SMILES encoders. The MCA mechanism enables to incorporate both positional information and long-range dependencies using a combination of convolutional layers and contextual attention mechanisms. Finally the encoded drug and cell-line features are concatenated and inputted to fully connected dense layers to predict the drug response.
Feature Representation
Cancer features:
Gene Expression: RMA-normalized
Drug features:
SMILES
Morgan fingerprints
URLs
References
1. M. Manica, et. al. “Toward Explainable Anticancer Compound Sensitivity Prediction via Multimodal Attention-Based Convolutional Encoders”, Mol. Pharmaceutics, 2019