Deep learning prediction of chemical-induced dose-dependent and context-specific multiplex phenotype responses and its application to personalized alzheimer’s disease drug repurposing
Table 2
The results of the ablation test.
The no-autoencoder training strategy indicates that neither the model is trained together with the autoencoder model nor the encoder parameters are initialized with the pre-training model. The pretraining-fine-tuning strategy means that the cell features encoder parameters are initialized with the pre-training model. The joint training strategy means that the autoencoder and supervised learning components are couples and jointly trained. We selected the Pearson correlation and Spearman correlation as the comparison metrics. We also selected the top 10, 20, 50, 100 upregulated and downregulated genes and evaluated the prediction precisions of them.