tCNNS

tCNNS is a software tool for phenotypic screening to predict the response of cancer cell lines to drugs. It utilizes two convolutional neural networks (CNNs) to extract features from the drug's simplified molecular input line entry specification (SMILES) format and the genetic feature vectors of cancer cell lines. The extracted features are fed into a fully connected network to predict the drug interaction and cancer cell lines.

Key points about tCNNS:

1. It achieves high accuracy in predicting drug effects on cancer cell lines, with a mean coefficient of determinant (R²) of 0.826 and a top quartile of 0.831 when the training and testing sets are divided based on the interaction pairs between drugs and cell lines.

2. The performance of tCNNS remains stable even with less but high-quality data and fewer features for the cancer cell lines.

3. It can handle outliers in the feature space effectively.

4. tCNNS provides insights into phenotypic screening, which can be valuable for anti-cancer drug discovery and re-purposing.

5. However, the performance of tCNNS drops in blind tests, indicating that further improvements may be necessary for its generalization ability.