Fig. 1From: Leveraging explainable AI for gut microbiome-based colorectal cancer classificationWaterfall plots of local explanations that correspond to four CRC subjects. At the bottom part of each plot, we can see the base value (0.53) which represents the expected value of the CRC class (the mean of model output over the training data). All SHAP values in each local explanation sum up to the predicted CRC probability of each subject (local additivity property of SHAP)Back to article page