A methodology for classifying tissue-specific metabolic and inflammatory receptor functions applied to subcutaneous and visceral adipose

Figures

• 

  Fig 1. Schematic workflow of data preprocessing and the proposed methodology.

• 

  Fig 2. Pathway enrichment analysis of ten representative WGCNA modules. The significantly enriched KEGG pathways (adjusted p-values < = 0.01) and enrichment scores calculated for subcutaneous and visceral adipose are presented. The columns present the modules (that include many labeled receptors) in each tissue and the rows represent the significantly enriched KEGG biological pathways. The matrix cells present the enrichment scores of the pathways for each module. The pathways are classified into six classes according to the BRITE classification and highlighted in the annotation column to the left (see the y-axis). For example, you can see that the modules that include many metabolic receptors (highlighted as “Metabolic” on the x-axis) are highly enriched with the “Metabolism” classification (colored in blue on the y-axis).

• 

  Fig 3. Top 10 features (pathways) and their average SHAP influence (absolute values) on the magnitude of the model prediction. A. Calculated for subcutaneous adipose. B. Calculated for visceral adipose. The SHAP method illustrates the magnitude of the effect of each feature (KEGG pathway) on each classification. The colored dots to the left highlight KEGG’s BRITE annotations of the pathways.

• 

  Fig 4. SHAP subcutaneous adipose “metabolic” class variable significance plot. The values show the impact of the feature on model output (prediction). The plot is composed of all receptors in the training data. SHAP values indicate how much the feature contributes to the classification.