Skip to main content
Fig. 4 | Journal of Translational Medicine

Fig. 4

From: The mechanistic functional landscape of retinitis pigmentosa: a machine learning-driven approach to therapeutic target discovery

Fig. 4

Influence and profile of KDTs on Retinitis Pigmentosa mechanistic map. A Heatmap of the 109 relevant predictive KDTs (X-axis) normalized (− 1, 1) SHAP scores over the stable circuits of the RP Map (Y-axis). The sign indicates the direction of the KDT influence over each specific circuit and the score value depicts how strong is the influence of a specific KDT for predicting the activity of a specific circuit. The top color bar represents the most frequent drug effects of that specific target (KDT). At the bottom, the KDT groups obtained by hierarchical clustering are represented by red (cluster 1 with 4 KDTs), yellow (Cluster 2 with 15 KDTs), and green (Cluster 3 with 90 KDTs) color. B) Barplot representing the distribution of SHAP scores showing the accumulated volume of positive (in red) and negative (in blue) scores (X-axis), for each specific KDT (Y-axis) overall RP Map. On the left side, KDTs are annotated with their gene symbol and colored by their cluster. C Chord diagram representing the shared influence of relevant KDTs over the circuits belonging to the 9 RP hallmarks modules. The size of the chord joining two hallmarks represents the number of shared KDTs, with their SHAP scores weighted by the % of the hallmark covered (no. of circuits tagged by the hallmark with the KDT/total number of circuits with the hallmark)

Back to article page