Introduction
Many genetic syndromes are associated with a distinctive facial gestalt
which can be used to expedite the diagnostic process. Although
high-throughput sequencing has helped to address the considerable
heterogeneity of many syndromes in a single test, the rare expertise of
dysmorphologists, which is still required for the data interpretation,
is often the bottleneck. In recent years, advances in machine learning
have enabled the development of NGP tools, that can be used to analyze
facial dysmorphology in patient portrait photos (Ferry et al., 2014;
Kuru et al., 2014; Gripp et al., 2016; Wang and Luo, 2016; Dudding-Byth
et al., 2017; Hadj-Rabia et al., 2017; Valentine et al., 2017; Liehr et
al., 2018; Gurovich et al., 2019; van der Donk et al., 2019; Hsieh et
al., 2022). Amongst them is GestaltMatcher, which is a deep
convolutional neural network that was trained on thousands of
molecularly confirmed cases and achieves high accuracies in the
identification of hundreds of syndromes (Hsieh et al., 2022). In this
paper, we describe how the results of this artificial intelligence
helped to solve a case with a typical phenotype of Koolen-de Vries
syndrome but an unusual disease-causing mutation.