Next generation phenotyping for diagnosis and phenotype-genotype correlations in Kabuki syndrome

Abstract

The field of dysmorphology has been changed by the use Artificial Intelligence (AI) and the development of Next Generation Phenotyping (NGP). The aim of this study was to propose a new NGP model for predicting KS (Kabuki Syndrome) on 2D facial photographs and distinguish KS1 (KS type 1, KMT2D-related) from KS2 (KS type 2, KDM6A-related). We included retrospectively and prospectively, from 1998 to 2023, all frontal and lateral pictures of patients with a molecular confirmation of KS. After automatic preprocessing, we extracted geometric and textural features. After incorporation of age, gender, and ethnicity, we used XGboost (eXtreme Gradient Boosting), a supervised machine learning classifier. The model was tested on an independent validation set. Finally, we compared the performances of our model with DeepGestalt (Face2Gene). The study included 1448 frontal and lateral facial photographs from 6 centers, corresponding to 634 patients (527 controls, 107 KS); 82 (78%) of KS patients had a variation in the KMT2D gene (KS1) and 23 (22%) in the KDM6A gene (KS2). We were able to distinguish KS from controls in the independent validation group with an accuracy of 95.8% (78.9-99.9%, p < 0.001) and distinguish KS1 from KS2 with an empirical Area Under the Curve (AUC) of 0.805 (0.729-0.880, p < 0.001). We report an automatic detection model for KS with high performances (AUC 0.993 and accuracy 95.8%). We were able to distinguish patients with KS1 from KS2, with an AUC of 0.805. These results outperform the current commercial AI-based solutions and expert clinicians.

Figure 1

Analysis pipeline, from the initial photograph to diagnostic probability. ROI, Region Of Interest; AAM, active appearance model; Faster RCNN, Faster Region-based Convolutional Neural Network; CLAHE, Contrast Limited Adaptative Histogram Equalization; GLCM, Gray-Level Co-occurrence Matrix; XGboost, eXtreme Gradient Boosting.

Figure 2

Average shapes in KS and controls and comparisons after Procrustes superimposition of frontal views, profile views, and external ears for three age groups. Blue = controls, Dark red = KS.

Figure 3

(A) Empirical ROC curves (training set) for KS with AUC in design №1. (B) ROC curves (validation set) for KS with AUC in design №1. AUC, area under the curve; KS, Kabuki Syndrome.

Figure 4

Classification using design №1 for proband 3 of the validation set. (A) and (B) Frontal and profile views of proband 3. (C) UMAP representation of the training data according to the two groups, with positioning of proband 3. (D) Histogram of predictions by the model. This child was also detected as KS by Face2Gene CLINIC. KS, Kabuki Syndrome.

Figure 5

Average shapes in KS1 and KS2 and comparisons after Procrustes superimposition of frontal views, lateral views, and external ears for three age groups. Orange = KS1, Dark red = KS2.

Figure 6

Empirical ROC curve (training set) for KS2 with AUC in design №2. AUC, Area Under the Curve; KS, Kabuki Syndrome.

Figure 7

Classification using design №2 for two probands of the training set. (A, B, E and F) Frontal and profile views of the two probands. (C and G) UMAP representations of the training data according to the two groups, with positioning of probands 3. (D and H) Histograms of predictions by the model. The phenotype included a reduced height of the midface, a thicker upper lip, and a vertical elongation of the external ear in the KS2 group (E and F). KS, Kabuki Syndrome.

Figure 8

Empirical ROC curve (training set) for KS1 PAV with AUC in design №3. AUC, Area Under the Curve; KS, Kabuki Syndrome; PAV, protein-altering variant.