Figure 2.

The conceptual framework diagram depicting machine learning in bacterial genome-wide association using extreme gradient boosting (XGBoost). Boosting is a technique of combining a set of weak classifiers or decision trees to increase prediction accuracy. Red dots represent an allelic variant, each grey bar represents a unique allele. Individual decision trees (1, 2, 3) fail to fully capture the allelic variants associated with the phenotype (e.g., extraintestinal abortion), but combining the trees together results in a process called boosting as it increases the discriminative power.