Extended Data Fig. 2: The major feature types contributed to the gene co-expression.
From: Deep learning decodes the principles of differential gene expression
We defined the gene pairs with the absolute Spearman’s correlation greater than 0.3 and the sign of the correlation matched with one with the ground truth as the successfully predicted gene pairs. The successfully predicted gene pairs of the model trained with the full set of features were split into three groups based on the performance of the models trained with only RNA features or promoter features.
