What this is
SHAP is the canonical name for Shapley additive explanations, a model-interpretation method that attributes a prediction to feature contributions. 1
Known alias: Shapley additive explanation. 1
No stable external identifier was reported for this method in the source document. 1
Use in the BacDive metal-tolerance study
SHAP feature importance was applied to the full XGBoost model to assess which predictors contributed to genome-based metal-tolerance scores. 1
Taxonomic class and order codes, together with n_metal_clusters, were among the top predictors in the SHAP ranking. 1
Phenotype features contributed minimally after taxonomy was included, supporting the interpretation that the measured phenotypes were primarily phylogenetic proxies rather than independent predictors of the composite metal-tolerance score. 1
The model was trained across 3,994 species and included 17 features in the full model. 1
This result supports environmental-resistome, where the full model reached R² = 0.633 after adding the number of metal-resistance gene clusters. 1
Use in ENIGMA genotype-to-phenotype prediction
In the ENIGMA study, correlation-grouped SHAP attributed 25.3% of total importance to a 63-feature genome-scale axis and 45.9% to condition class; membrane adaptation, tRNA modification, aromatic catabolism, and flagellar-motility KO blocks each contributed approximately 2%. 2
This refines the BacDive result: SHAP can expose broad taxonomic or genome-scale proxies, but the larger ENIGMA corpus also identified condition-specific predictors, including K03762 (proP), K10440 (rbsC), K01857 (pcaB), K13633 (ftrA), and K01214 (treX). 2
The ENIGMA analysis further used SHAP to compare predicted growth with Fitness Browser fitness effects; 18.7% of top SHAP KOs expanded into correlation-grouped gene blocks had significant matched-condition fitness effects versus a 16.3% random baseline, a 1.19× enrichment. This supports interpreting feature-attribution results as potentially mechanistic while refining the distinction between gene presence across genera and gene essentiality within one strain. 2
The detailed ENIGMA application is described in Genotype × Condition to Phenotype Prediction from ENIGMA Growth Curves. 2
Related material
The method is described in the study summary BacDive Phenotype Signatures of Metal Tolerance. 1
Its interpretation also relates to pangenome-integration, because the analysis integrated BacDive phenotypes with species-level pangenome and metal-score data. 1