Fig. 6From: LANDMark: an ensemble approach to the supervised selection of biomarkers in high-throughput sequencing dataThe top 20 ASVs, selected using recursive feature elimination and LANDMark, from the F230 subset. These ASVs are used by LANDMark to help identify the Athabasca or Peace River Delta. The KernelSHAP method was used to calculate the SHAP values for each ASV in each sample. Each point is a sample and the color of each point reflects the presence (pink) or absence (blue) of the ASV listed along the y-axis. The higher the absolute value of a sample’s score for a particular ASV along the x-axis, the more strongly that ASV shifts the prediction of a sample. Positive SHAP values push the prediction towards the Athabasca River Delta; negative SHAP values push prediction towards the Peace RiverBack to article page