Fig. 5

Interpretable machine learning models uncover context-dependent interactions. A–F The weights of a linear regression model predicting the essentiality of FAM50A trained on cell lines from the specified tissues. Red bars represent positive weights while blue bars represent negative weights. G The features of a decision tree learned using the sci-kit learn package for predicting FAM50A with a maximum depth 4. The decision tree predictions work by traversing the tree either left or right depending on whether the expression of the gene in the node is less than or equal to the value in the node (left) or not (right). The leaves represent predictions, or the average essentiality of FAM50A in the train data for cell lines that satisfy the conditions of reaching that leaf node