GestaltMatcher: Overcoming the limits of rare disease matching using facial phenotypic descriptors

Abstract

The majority of monogenic disorders cause craniofacial abnormalities with characteristic facial morphology. These disorders can be diagnosed more efficiently with the support of computer-aided next-generation phenotyping tools, such as DeepGestalt. These tools have learned to associate facial phenotypes with the underlying syndrome through training on thousands of patient photographs. However, this “supervised” approach means that diagnoses are only possible if they were part of the training set. To improve recognition of ultra-rare diseases, we created GestaltMatcher, which uses a deep convolutional neural network based on the DeepGestalt framework. We used photographs of 21,836 patients with 1,362 rare disorders to define a “Clinical Face Phenotype Space”. Distance between cases in the phenotype space defines syndromic similarity, allowing test patients to be matched to a molecular diagnosis even when the disorder was not included in the training set. Similarities among patients with previously unknown disease genes can also be detected. Therefore, in concert with mutation data, GestaltMatcher could accelerate the clinical diagnosis of patients with ultra-rare disorders and facial dysmorphism.

De novo variants in thePSMC3proteasome AAA-ATPase subunit gene cause neurodevelopmental disorders associated with type I interferonopathies

“…Most children (17/19; 89%) displayed dysmorphic facial features, including notably tall or broad forehead (7/19; 37%), thin upper lip with down-turned corners of mouth (6/19; 32%), abnormal palate (5/19; 265/19; 26%), epicanthal folds (5/19; 26%), and orofacial clefts (2/19; 10%). Computational analysis of facial morphology by GestaltMatcher [42] revealed that facial dysmorphism among the PSMC3 subjects was rather heterogeneous with similarities only observed between patients carrying identical variants (Figure S2).…”

“…One major phenotype hallmark of all individuals with PSMC3 variants is the predominance of neurodevelopmental or neuropsychiatric symptoms ( [42] revealed that facial dysmorphism among the PSMC3 subjects was rather heterogeneous with similarities only observed between patients carrying identical variants (Figure S2).…”

KBG Syndrome: Prospective Videoconferencing and Use of AI-driven Facial Phenotyping in 25 New Patients

“…Our focus was on the clinical and phenotypic characteristics of KBG syndrome as presented in our cohort. GestaltMatcher spans a 320-dimensional clinical face phenotype space (CFPS) defined by the feature vectors derived from DeepGestalt [6].…”

“…met and interviewed twenty-five individuals (11 females, 14 males) from 22 families with KBG syndrome, to investigate further the role of epilepsy and other conditions in possibly affecting the trajectory of neurodevelopment in these individuals. In addition, it was asked whether the current state of facial recognition software can play a role in the diagnosis of this syndrome, so efforts were made to test the current status of two leading algorithms in the field [5,6]. In an ideal world, a facial photograph could be combined with medical records and variant prioritization efforts, after exome or whole genome sequencing, to more accurately classify new missense and other variants in rare syndromes as likely pathogenic [7].…”

Neural networks for classification and image generation of aging in genetic syndromes

“…Neural networks are emerging as powerful tools in many areas of biomedical research and are starting to impact clinical care. In the field of genomics, these methods are applied in multiple ways, including generating differential diagnoses for patients with possible genetic syndrome based on images, [1][2][3][4] analysis of DNA sequencing data 5 (including phenotype-based annotation 6 and variant classification 7 ), and prediction of protein structure. 8; 9 In clinical genetics, the rarity, complexity, and number of different diseases, 10; 11 coupled with a dearth of trained experts, 12; 13 can lead to delayed diagnosis and suboptimal management.…”