Advances in computer-assisted syndrome recognition by the example of inborn errors of metabolism

Facial recognition models for identifying genetic syndromes associated with pulmonary stenosis in children

BACKGROUND

Williams-Beuren syndrome, Noonan syndrome, and Alagille syndrome are common types of genetic syndromes (GSs) characterized by distinct facial features, pulmonary stenosis, and delayed growth. In clinical practice, differentiating these three GSs remains a challenge. Facial gestalts serve as a diagnostic tool for recognizing Williams-Beuren syndrome, Noonan syndrome, and Alagille syndrome. Pretrained foundation models (PFMs) can be considered the foundation for small-scale tasks. By pretraining with a foundation model, we propose facial recognition models for identifying these syndromes.

METHODS

A total of 3297 (n = 1666) facial photos were obtained from children diagnosed with Williams-Beuren syndrome (n = 174), Noonan syndrome (n = 235), and Alagille syndrome (n = 51), and from children without GSs (n = 1206). The photos were randomly divided into five subsets, with each syndrome and non-GS equally and randomly distributed in each subset. The proportion of the training set and the test set was 4:1. The ResNet-100 architecture was employed as the backbone model. By pretraining with a foundation model, we constructed two face recognition models: one utilizing the ArcFace loss function, and the other employing the CosFace loss function. Additionally, we developed two models using the same architecture and loss function but without pretraining. The accuracy, precision, recall, and F1 score of each model were evaluated. Finally, we compared the performance of the facial recognition models to that of five pediatricians.

RESULTS

Among the four models, ResNet-100 with a PFM and CosFace loss function achieved the best accuracy (84.8%). Of the same loss function, the performance of the PFMs significantly improved (from 78.5% to 84.5% for the ArcFace loss function, and from 79.8% to 84.8% for the CosFace loss function). With and without the PFM, the performance of the CosFace loss function models was similar to that of the ArcFace loss function models (79.8% vs 78.5% without PFM; 84.8% vs 84.5% with PFM). Among the five pediatricians, the highest accuracy (0.700) was achieved by the senior-most pediatrician with genetics training. The accuracy and F1 scores of the pediatricians were generally lower than those of the models.

CONCLUSIONS

A facial recognition-based model has the potential to improve the identification of three common GSs with pulmonary stenosis. PFMs might be valuable for building screening models for facial recognition. Key messages What is already known on this topic:  Early identification of genetic syndromes (GSs) is crucial for the management and prognosis of children with pulmonary stenosis (PS). Facial phenotyping with convolutional neural networks (CNNs) often requires large-scale training data, limiting its usefulness for GSs. What this study adds:  We successfully built multi-classification models based on face recognition using a CNN to accurately identify three common PS-associated GSs. ResNet-100 with a pretrained foundation model (PFM) and CosFace loss function achieved the best accuracy (84.8%). Pretrained with the foundation model, the performance of the models significantly improved, although the impact of the type of loss function appeared to be minimal. How this study might affect research, practice, or policy:  A facial recognition-based model has the potential to improve the identification of GSs in children with PS. The PFM might be valuable for building identification models for facial detection.

Loss-of-Function Variants in CUL3 Cause a Syndromic Neurodevelopmental Disorder

OBJECTIVE

De novo variants in cullin-3 ubiquitin ligase (CUL3) have been strongly associated with neurodevelopmental disorders (NDDs), but no large case series have been reported so far. Here, we aimed to collect sporadic cases carrying rare variants in CUL3, describe the genotype-phenotype correlation, and investigate the underlying pathogenic mechanism.

METHODS

Genetic data and detailed clinical records were collected via multicenter collaboration. Dysmorphic facial features were analyzed using GestaltMatcher. Variant effects on CUL3 protein stability were assessed using patient-derived T-cells.

RESULTS

We assembled a cohort of 37 individuals with heterozygous CUL3 variants presenting a syndromic NDD characterized by intellectual disability with or without autistic features. Of these, 35 have loss-of-function (LoF) and 2 have missense variants. CUL3 LoF variants in patients may affect protein stability leading to perturbations in protein homeostasis, as evidenced by decreased ubiquitin-protein conjugates in vitro. Notably, we show that 4E-BP1 (EIF4EBP1), a prominent substrate of CUL3, fails to be targeted for proteasomal degradation in patient-derived cells.

INTERPRETATION

Our study further refines the clinical and mutational spectrum of CUL3-associated NDDs, expands the spectrum of cullin RING E3 ligase-associated neuropsychiatric disorders, and suggests haploinsufficiency via LoF variants is the predominant pathogenic mechanism. ANN NEUROL 2024.

RNA analysis and computer-aided facial phenotyping help to classify a novel TRIO splice site variant

Pathogenic variants in TRIO, encoding the guanine nucleotide exchange factor, are associated with two distinct neurodevelopmental delay phenotypes: gain-of-function missense mutations within the spectrin repeats are causative for a severe developmental delay with macrocephaly (MIM: 618825), whereas loss-of-function missense variants in the GEF1 domain and truncating variants throughout the gene lead to a milder developmental delay and microcephaly (MIM: 617061). In three affected family members with mild intellectual disability/NDD and microcephaly, we detected a novel heterozygous TRIO variant at the last coding base of exon 31 (NM_007118.4:c.4716G>A). RNA analysis from patient-derived lymphoblastoid cells confirmed aberrant splicing resulting in the skipping of exon 31 (r.4615_4716del), leading to an in-frame deletion in the first Pleckstrin homology subdomain of the GEF1 domain: p.(Thr1539_Lys1572del). To test for a distinct gestalt, facial characteristics of the family members and 41 previously published TRIO cases were systematically evaluated via GestaltMatcher. Computational analysis of the facial gestalt suggests a distinguishable facial TRIO-phenotype not outlined in the existing literature.

Validation of 3 Computer-Aided Facial Phenotyping Tools (DeepGestalt, GestaltMatcher, and D-Score): Comparative Diagnostic Accuracy Study

BACKGROUND

While characteristic facial features provide important clues for finding the correct diagnosis in genetic syndromes, valid assessment can be challenging. The next-generation phenotyping algorithm DeepGestalt analyzes patient images and provides syndrome suggestions. GestaltMatcher matches patient images with similar facial features. The new D-Score provides a score for the degree of facial dysmorphism.

OBJECTIVE

We aimed to test state-of-the-art facial phenotyping tools by benchmarking GestaltMatcher and D-Score and comparing them to DeepGestalt.

METHODS

Using a retrospective sample of 4796 images of patients with 486 different genetic syndromes (London Medical Database, GestaltMatcher Database, and literature images) and 323 inconspicuous control images, we determined the clinical use of D-Score, GestaltMatcher, and DeepGestalt, evaluating sensitivity; specificity; accuracy; the number of supported diagnoses; and potential biases such as age, sex, and ethnicity.

RESULTS

DeepGestalt suggested 340 distinct syndromes and GestaltMatcher suggested 1128 syndromes. The top-30 sensitivity was higher for DeepGestalt (88%, SD 18%) than for GestaltMatcher (76%, SD 26%). DeepGestalt generally assigned lower scores but provided higher scores for patient images than for inconspicuous control images, thus allowing the 2 cohorts to be separated with an area under the receiver operating characteristic curve (AUROC) of 0.73. GestaltMatcher could not separate the 2 classes (AUROC 0.55). Trained for this purpose, D-Score achieved the highest discriminatory power (AUROC 0.86). D-Score's levels increased with the age of the depicted individuals. Male individuals yielded higher D-scores than female individuals. Ethnicity did not appear to influence D-scores.

CONCLUSIONS

If used with caution, algorithms such as D-score could help clinicians with constrained resources or limited experience in syndromology to decide whether a patient needs further genetic evaluation. Algorithms such as DeepGestalt could support diagnosing rather common genetic syndromes with facial abnormalities, whereas algorithms such as GestaltMatcher could suggest rare diagnoses that are unknown to the clinician in patients with a characteristic, dysmorphic face.

Loss-of-function variants in CUL3 cause a syndromic neurodevelopmental disorder

Purpose

De novo variants in CUL3 (Cullin-3 ubiquitin ligase) have been strongly associated with neurodevelopmental disorders (NDDs), but no large case series have been reported so far. Here we aimed to collect sporadic cases carrying rare variants in CUL3, describe the genotype-phenotype correlation, and investigate the underlying pathogenic mechanism.

Methods

Genetic data and detailed clinical records were collected via multi-center collaboration. Dysmorphic facial features were analyzed using GestaltMatcher. Variant effects on CUL3 protein stability were assessed using patient-derived T-cells.

Results

We assembled a cohort of 35 individuals with heterozygous CUL3 variants presenting a syndromic NDD characterized by intellectual disability with or without autistic features. Of these, 33 have loss-of-function (LoF) and two have missense variants. CUL3 LoF variants in patients may affect protein stability leading to perturbations in protein homeostasis, as evidenced by decreased ubiquitin-protein conjugates in vitro . Specifically, we show that cyclin E1 (CCNE1) and 4E-BP1 (EIF4EBP1), two prominent substrates of CUL3, fail to be targeted for proteasomal degradation in patient-derived cells.

Conclusion

Our study further refines the clinical and mutational spectrum of CUL3 -associated NDDs, expands the spectrum of cullin RING E3 ligase-associated neuropsychiatric disorders, and suggests haploinsufficiency via LoF variants is the predominant pathogenic mechanism.

3D facial analysis for rare disease diagnosis and treatment monitoring: Proof-Of-Concept plan for hereditary angioedema

Rare diseases pose a diagnostic conundrum to even the most experienced clinicians around the world. The technology could play an assistive role in hastening the diagnosis process. Data-driven methodologies can identify distinctive disease features and create a definitive diagnostic spectrum. The healthcare professionals in developed and developing nations would benefit immensely from these approaches resulting in quicker diagnosis and enabling early care for the patients. Hereditary Angioedema is one such rare disease that requires a lengthy diagnostic cascade ensuing massive patient inconvenience and cost burden on the healthcare system. It is hypothesized that facial analysis with advanced imaging and algorithmic association can create an ideal diagnostic peer to the clinician while assimilating signs and symptoms in the hospital. 3D photogrammetry has been applied to diagnose rare diseases in various cohorts. The facial features are captured at a granular level in utmost finer detail. A validated and proven algorithm-powered software provides recommendations in real-time. Thus, paving the way for quick and early diagnosis to well-trained or less trained clinicians in different settings around the globe. The generated evidence indicates the strong applicability of 3 D photogrammetry in association with proprietary Cliniface software to Hereditary Angioedema for aiding in the diagnostic process. The approach, mechanism, and beneficial impact have been sketched out appropriately herein. This blueprint for hereditary angioedema may have far-reaching consequences beyond disease diagnosis to benefit all the stakeholders in the healthcare arena including research and new drug development.

Case Report: The success of face analysis technology in extremely rare genetic diseases in Korea: Tatton–Brown–Rahman syndrome and Say–Barber –Biesecker–Young–Simpson variant of ohdo syndrome

Tatton–Brown–Rahman syndrome (TBRS) and Say–Barber–Biesecker– Young–Simpson variant of Ohdo syndrome (SBBYSS) are extremely rare genetic disorders with less than 100 reported cases. Patients with these disorders exhibit a characteristic facial dysmorphism: TBRS is characterized by a round face, a straight and thick eyebrow, and prominent maxillary incisors, whereas SBBYSS is characterized by mask-like facies, blepharophimosis, and ptosis. The usefulness of Face2Gene as a tool for the identification of dysmorphology syndromes is discussed, because, in these patients, it suggested TBRS and SBBYSS within the top five candidate disorders. Face2Gene is useful for the diagnosis of extremely rare diseases in Korean patients, suggesting the possibility of expanding its clinical applications.

Evaluation of a computer-based facial dysmorphology analysis algorithm (Face2Gene) using standardized textbook photos

BACKGROUND

Genetic syndromes often have ocular involvement. Ophthalmologists may have difficulty identifying dysmorphic features in genetic syndrome evaluations. We investigated the sensitivity and specificity of Face2Gene (F2G), a digital image analysis software trained on integrating dysmorphic features, by analysing patient photos from genetics textbooks.

METHODS

We analysed all clear facial photos contained within the textbooks Smith's Recognizable Patterns of Human Malformation and Genetic Diseases of the Eye using F2G under standard lighting conditions. Variables captured include colour versus grey scale photo, the gender of the patient (if known), age of the patient (if known), disease categories, diagnosis as listed in the textbook, and whether the disease has ophthalmic involvement (as described in the textbook entries). Any photos rejected by F2G were excluded. We analysed the data for accuracy, sensitivity, and specificity based on disease categories as outlined in Smith's Recognizable Patterns of Malformation.

RESULTS

We analysed 353 photos found within two textbooks. The exact book diagnosis was identified by F2G in 150 (42.5%) entries, and was included in the top three differential diagnoses in 191 (54.1%) entries. F2G is highly sensitive for craniosynostosis syndromes (point estimate [PE] 80.0%, 95% confidence interval [CI] 56.3-94.3%, P = 0.0118) and syndromes with facial defects as a major feature (PE 77.8%, 95% CI 52.4-93.6%, P = 0.0309). F2G was highly specific (PE > 83percentage with P < 0.001) for all disease categories.

CONCLUSIONS

F2G is a useful tool for paediatric ophthalmologists to help build a differential diagnosis when evaluating children with dysmorphic facial features.

Using deep-neural-network-driven facial recognition to identify distinct Kabuki syndrome 1 and 2 gestalt

Kabuki syndrome (KS) is a rare genetic disorder caused by mutations in two major genes, KMT2D and KDM6A, that are responsible for Kabuki syndrome 1 (KS1, OMIM147920) and Kabuki syndrome 2 (KS2, OMIM300867), respectively. We lack a description of clinical signs to distinguish KS1 and KS2. We used facial morphology analysis to detect any facial morphological differences between the two KS types. We used a facial-recognition algorithm to explore any facial morphologic differences between the two types of KS. We compared several image series of KS1 and KS2 individuals, then compared images of those of Caucasian origin only (12 individuals for each gene) because this was the main ethnicity in this series. We also collected 32 images from the literature to amass a large series. We externally validated results obtained by the algorithm with evaluations by trained clinical geneticists using the same set of pictures. Use of the algorithm revealed a statistically significant difference between each group for our series of images, demonstrating a different facial morphotype between KS1 and KS2 individuals (mean area under the receiver operating characteristic curve = 0.85 [p = 0.027] between KS1 and KS2). The algorithm was better at discriminating between the two types of KS with images from our series than those from the literature (p = 0.0007). Clinical geneticists trained to distinguished KS1 and KS2 significantly recognised a unique facial morphotype, which validated algorithm findings (p = 1.6e-11). Our deep-neural-network-driven facial-recognition algorithm can reveal specific composite gestalt images for KS1 and KS2 individuals.

Investigation of a dysmorphic facial phenotype in patients with Gaucher disease types 2 and 3

Gaucher disease (GD) is a rare lysosomal storage disorder that is divided into three subtypes based on presentation of neurological manifestations. Distinguishing between the types has important implications for treatment and counseling. Yet, patients with neuronopathic forms of GD, types 2 and 3, often present at young ages and can have overlapping phenotypes. It has been shown that new technologies employing artificial intelligence and facial recognition software can assist with dysmorphology assessments. Though classically not associated nor previously described with a dysmorphic facial phenotype, this study investigated whether a facial recognition platform could distinguish between photos of patients with GD2 and GD3 and discriminate between them and photos of healthy controls. Each cohort included over 100 photos. A cross validation scheme including a series of binary comparisons between groups was used. Outputs included a composite photo of each cohort and either a receiver operating characteristic curve or a confusion matrix. Binary comparisons showed that the software could correctly group photos at least 89% of the time. Multiclass comparison between GD2, GD3, and healthy controls demonstrated a mean accuracy of 76.6%, compared to a 37.7% chance for random comparison. Both GD2 and GD3 have now been added to the facial recognition platform as established syndromes that can be identified by the algorithm. These results suggest that facial recognition and artificial intelligence, though no substitute for other diagnostic methods, may aid in the recognition of neuronopathic GD. The algorithm, in concert with other clinical features, also appears to distinguish between young patients with GD2 and GD3, suggesting that this tool can help facilitate earlier implementation of appropriate management.

Facial recognition accuracy in photographs of Thai neonates with Down syndrome among physicians and the Face2Gene application

Down syndrome (DS) is typically recognizable in those who present with multiple dysmorphism, especially in regard to facial phenotypes. However, as the presentation of DS in neonates is less obvious, a phenotype-based presumptive diagnosis is more challenging. Recently, an artificial intelligence (AI) application, Face2Gene, was developed to help physicians recognize specific genetic syndromes by using two-dimensional facial photos. As of yet, there has not been any study comparing accuracy among physicians or applications. Our objective was to compare the facial recognition accuracy of DS in Thai neonates, using facial photographs, among physicians and the Face2Gene. Sixty-four Thai neonates at Thammasat University Hospital, with genetic testing and signed parental consent, were divided into a DS group (25) and non-DS group (39). Non-DS was further divided into unaffected (19) and those affected with other syndromes (20). Our results revealed physician accuracy (89%) was higher than the Face2Gene (81%); however, the application was higher in sensitivity (100%) than physicians (86%). While this application can serve as a helpful assistant in facilitating any genetic syndrome such as DS, to aid clinicians in recognizing DS facial features in neonates, it is not a replacement for well-trained doctors.

SRD5A3-CDG: 3D structure modeling, clinical spectrum, and computer-based dysmorphic facial recognition

Pathogenic variants in Steroid 5 alpha reductase type 3 (SRD5A3) cause rare inherited congenital disorder of glycosylation known as SRD5A3-CDG (MIM# 612379). To date, 43 affected individuals have been reported. Despite the development of various dysmorphic features in significant number of patients, facial recognition entity has not yet been established for SRD5A3-CDG. Herein, we reported a novel SRD5A3 missense pathogenic variant c.460 T > C p.(Ser154Pro). The 3D structural modeling of the SRD5A3 protein revealed additional transmembrane α-helices and predicted that the p.(Ser154Pro) variant is located in a potential active site and is capable of reducing its catalytic efficiency. Based on phenotypes of our patients and all published SRD5A3-CDG cases, we identified the most common clinical features as well as some recurrent dysmorphic features such as arched eyebrows, wide eyes, shallow nasal bridge, short nose, and large mouth. Based on facial digital 2D images, we successfully designed and validated a SRD5A3-CDG computer based dysmorphic facial analysis, which achieved 92.5% accuracy. The current work integrates genotypic, 3D structural modeling and phenotypic characteristics of CDG-SRD5A3 cases with the successful development of computer tool for accurate facial recognition of CDG-SRD5A3 complex cases to assist in the diagnosis of this particular disorder globally.

Efficiency of Computer-Aided Facial Phenotyping (DeepGestalt) in Individuals With and Without a Genetic Syndrome: Diagnostic Accuracy Study

Background

Collectively, an estimated 5% of the population have a genetic disease. Many of them feature characteristics that can be detected by facial phenotyping. Face2Gene CLINIC is an online app for facial phenotyping of patients with genetic syndromes. DeepGestalt, the neural network driving Face2Gene, automatically prioritizes syndrome suggestions based on ordinary patient photographs, potentially improving the diagnostic process. Hitherto, studies on DeepGestalt’s quality highlighted its sensitivity in syndromic patients. However, determining the accuracy of a diagnostic methodology also requires testing of negative controls.

Objective

The aim of this study was to evaluate DeepGestalt's accuracy with photos of individuals with and without a genetic syndrome. Moreover, we aimed to propose a machine learning–based framework for the automated differentiation of DeepGestalt’s output on such images.

Methods

Frontal facial images of individuals with a diagnosis of a genetic syndrome (established clinically or molecularly) from a convenience sample were reanalyzed. Each photo was matched by age, sex, and ethnicity to a picture featuring an individual without a genetic syndrome. Absence of a facial gestalt suggestive of a genetic syndrome was determined by physicians working in medical genetics. Photos were selected from online reports or were taken by us for the purpose of this study. Facial phenotype was analyzed by DeepGestalt version 19.1.7, accessed via Face2Gene CLINIC. Furthermore, we designed linear support vector machines (SVMs) using Python 3.7 to automatically differentiate between the 2 classes of photographs based on DeepGestalt's result lists.

Results

We included photos of 323 patients diagnosed with 17 different genetic syndromes and matched those with an equal number of facial images without a genetic syndrome, analyzing a total of 646 pictures. We confirm DeepGestalt’s high sensitivity (top 10 sensitivity: 295/323, 91%). DeepGestalt’s syndrome suggestions in individuals without a craniofacially dysmorphic syndrome followed a nonrandom distribution. A total of 17 syndromes appeared in the top 30 suggestions of more than 50% of nondysmorphic images. DeepGestalt’s top scores differed between the syndromic and control images (area under the receiver operating characteristic [AUROC] curve 0.72, 95% CI 0.68-0.76; P<.001). A linear SVM running on DeepGestalt’s result vectors showed stronger differences (AUROC 0.89, 95% CI 0.87-0.92; P<.001).

Conclusions

DeepGestalt fairly separates images of individuals with and without a genetic syndrome. This separation can be significantly improved by SVMs running on top of DeepGestalt, thus supporting the diagnostic process of patients with a genetic syndrome. Our findings facilitate the critical interpretation of DeepGestalt’s results and may help enhance it and similar computer-aided facial phenotyping tools.

Differentiating molecular etiologies of Angelman syndrome through facial phenotyping using deep learning

Angelman syndrome (AS) is caused by several genetic mechanisms that impair the expression of maternally-inherited UBE3A through deletions, paternal uniparental disomy (UPD), UBE3A pathogenic variants, or imprinting defects. Current methods of differentiating the etiology require molecular testing, which is sometimes difficult to obtain. Recently, computer-based facial analysis systems have been used to assist in identifying genetic conditions based on facial phenotypes. We sought to understand if the facial-recognition system DeepGestalt could find differences in phenotype between molecular subtypes of AS. Images and molecular data on 261 individuals with AS ranging from 10 months through 32 years were analyzed by DeepGestalt in a cross-validation model with receiver operating characteristic (ROC) curves generated. The area under the curve (AUC) of the ROC for each molecular subtype was compared and ranked from least to greatest differentiable phenotype. We determined that DeepGestalt demonstrated a high degree of discrimination between the deletion subtype and UPD or imprinting defects, and a lower degree of discrimination with the UBE3A pathogenic variants subtype. Our findings suggest that DeepGestalt can recognize subclinical differences in phenotype based on etiology and may provide decision support for testing.

Automated syndrome diagnosis by three-dimensional facial imaging

Purpose

Deep phenotyping is an emerging trend in precision medicine for genetic disease. The shape of the face is affected in 30–40% of known genetic syndromes. Here, we determine whether syndromes can be diagnosed from 3D images of human faces.

Methods

We analyzed variation in three-dimensional (3D) facial images of 7057 subjects: 3327 with 396 different syndromes, 727 of their relatives, and 3003 unrelated, unaffected subjects. We developed and tested machine learning and parametric approaches to automated syndrome diagnosis using 3D facial images.

Results

Unrelated, unaffected subjects were correctly classified with 96% accuracy. Considering both syndromic and unrelated, unaffected subjects together, balanced accuracy was 73% and mean sensitivity 49%. Excluding unrelated, unaffected subjects substantially improved both balanced accuracy (78.1%) and sensitivity (56.9%) of syndrome diagnosis. The best predictors of classification accuracy were phenotypic severity and facial distinctiveness of syndromes. Surprisingly, unaffected relatives of syndromic subjects were frequently classified as syndromic, often to the syndrome of their affected relative.

Conclusion

Deep phenotyping by quantitative 3D facial imaging has considerable potential to facilitate syndrome diagnosis. Furthermore, 3D facial imaging of “unaffected” relatives may identify unrecognized cases or may reveal novel examples of semidominant inheritance.

Shared facial phenotype of patients with mucolipidosis type IV: A clinical observation reaffirmed by next generation phenotyping

BACKGROUND

Mucolipidosis type IV (ML-IV) is a rare autosomal-recessive lysosomal storage disease, caused by mutations in MCOLN1. ML-IV manifests with developmental delay, esotropia and corneal clouding. While the clinical phenotype is well-described, the diagnosis of ML-IV is often challenging and elusive.

OBJECTIVE

Our experience with ML-IV patients brought to the clinical observation that they share common and identifiable facial features, not yet described in the literature to date. Here, we utilized a computerized facial analysis tool to establish this association.

METHODS

Using the DeepGestalt algorithm, 50 two-dimensional facial images of ten ML-IV patients were analyzed, and compared to unaffected controls (n = 98) and to individuals affected with other genetic disorders (n = 99). Results were expressed in terms of the area-under-the-curve (AUC) of the receiver-operating-characteristic curve (ROC).

RESULTS

When compared to unaffected cases and to cases diagnosed with syndromes other than ML-IV, the ML-IV cohort showed an AUC of 0.822 (p value < 0.01) and an AUC of 0.885 (p value < 0.001), respectively.

CONCLUSIONS

We describe recognizable facial features typical in patients with ML-IV. Reaffirmed by the DeepGestalt technology, the described common facial phenotype adds to the tools currently available for clinicians and may thus assist in reaching an earlier diagnosis of this rare and underdiagnosed disorder.

Clinical versus automated assessments of morphological variants in twins with and without neurodevelopmental disorders

Physical examinations are recommended as part of a comprehensive evaluation for individuals with neurodevelopmental disorders (NDDs), such as autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder. These examinations should include assessment for morphological variants. Previous studies have shown an increase in morphological variants in individuals with NDDs, particularly ASD, and that these variants may be present in greater amounts in individuals with genetic alterations. Unfortunately, assessment for morphological variants can be subjective and time-consuming, and require a high degree of clinical expertise. Therefore, objective, automated methods of morphological assessment are desirable. This study compared the use of Face2Gene, an automated tool to explore facial morphological variants, to clinical consensus assessment, using a cohort of N = 290 twins enriched for NDDs (n = 135 with NDD diagnoses). Agreement between automated and clinical assessments were satisfactory to complete (78.3-100%). In our twin sample, individuals with NDDs did not have greater numbers of facial morphological variants when compared to those with typical development, nor when controlling for shared genetic and environmental factors within twin pairs. Common facial morphological variants in those with and without NDDs were similar and included thick upper lip vermilion, abnormality of the nasal tip, long face, and upslanted palpebral fissure. We conclude that although facial morphological variants can be assessed reliably in NDDs with automated tools like Face2Gene, clinical utility is limited when just exploring the facial region. Therefore, currently, automated assessments may best complement, rather than replace, in-person clinical assessments.

Evaluating Face2Gene as a Tool to Identify Cornelia de Lange Syndrome by Facial Phenotypes

Characteristic or classic phenotype of Cornelia de Lange syndrome (CdLS) is associated with a recognisable facial pattern. However, the heterogeneity in causal genes and the presence of overlapping syndromes have made it increasingly difficult to diagnose only by clinical features. DeepGestalt technology, and its app Face2Gene, is having a growing impact on the diagnosis and management of genetic diseases by analysing the features of affected individuals. Here, we performed a phenotypic study on a cohort of 49 individuals harbouring causative variants in known CdLS genes in order to evaluate Face2Gene utility and sensitivity in the clinical diagnosis of CdLS. Based on the profile images of patients, a diagnosis of CdLS was within the top five predicted syndromes for 97.9% of our cases and even listed as first prediction for 83.7%. The age of patients did not seem to affect the prediction accuracy, whereas our results indicate a correlation between the clinical score and affected genes. Furthermore, each gene presents a different pattern recognition that may be used to develop new neural networks with the goal of separating different genetic subtypes in CdLS. Overall, we conclude that computer-assisted image analysis based on deep learning could support the clinical diagnosis of CdLS.

Success of Face Analysis Technology in Rare Genetic Diseases Diagnosed by Whole-Exome Sequencing: A Single-Center Experience

The diagnosis of rare genetic diseases is one of the most difficult areas in medicine. Whole-exome sequencing (WES) technology makes it easier to diagnose these diseases. In addition, next-generation phenotyping can help to diagnose computer-based algorithms. Detailed dysmorphologic findings of 25 patients diagnosed by WES in our center were described. The success of this technology in diagnosing rare genetic diseases was investigated by scanning the photographs of 25 patients with Face2Gene application. The application listed possible preliminary diagnoses (30 disease suggestion). Of these, 12 (48%) cases were correctly matched. The most common disease group in the patients was neurological disease (96%). The most common mode of inheritance in the patients was autosomal recessive. The rate of consanguineous marriages was determined in 80% of the patients. Ten patients had microcephaly and 7 patients had corpus callosum anomaly. In our study, we found that the success of Face2Gene was lower than described in the literature. We think that the probable cause of this condition is that the cases are very rare, and there is not enough data about these diseases in the application. Therefore, it is recommended that applications should be used more frequently by pediatricians and clinical geneticists. The diagnosis of rare diseases still is quite difficult. Nowadays, WES is a successful method. However, applications such as Face2Gene help to make a clinical prediagnosis and create a larger database.

Mutations in PIGU Impair the Function of the GPI Transamidase Complex, Causing Severe Intellectual Disability, Epilepsy, and Brain Anomalies

The glycosylphosphatidylinositol (GPI) anchor links over 150 proteins to the cell surface and is present on every cell type. Many of these proteins play crucial roles in neuronal development and function. Mutations in 18 of the 29 genes implicated in the biosynthesis of the GPI anchor have been identified as the cause of GPI biosynthesis deficiencies (GPIBDs) in humans. GPIBDs are associated with intellectual disability and seizures as their cardinal features. An essential component of the GPI transamidase complex is PIGU, along with PIGK, PIGS, PIGT, and GPAA1, all of which link GPI-anchored proteins (GPI-APs) onto the GPI anchor in the endoplasmic reticulum (ER). Here, we report two homozygous missense mutations (c.209T>A [p.Ile70Lys] and c.1149C>A [p.Asn383Lys]) in five individuals from three unrelated families. All individuals presented with global developmental delay, severe-to-profound intellectual disability, muscular hypotonia, seizures, brain anomalies, scoliosis, and mild facial dysmorphism. Using multicolor flow cytometry, we determined a characteristic profile for GPI transamidase deficiency. On granulocytes this profile consisted of reduced cell-surface expression of fluorescein-labeled proaerolysin (FLAER), CD16, and CD24, but not of CD55 and CD59; additionally, B cells showed an increased expression of free GPI anchors determined by T5 antibody. Moreover, computer-assisted facial analysis of different GPIBDs revealed a characteristic facial gestalt shared among individuals with mutations in PIGU and GPAA1. Our findings improve our understanding of the role of the GPI transamidase complex in the development of nervous and skeletal systems and expand the clinical spectrum of disorders belonging to the group of inherited GPI-anchor deficiencies.

Enabling Global Clinical Collaborations on Identifiable Patient Data: The Minerva Initiative

The clinical utility of computational phenotyping for both genetic and rare diseases is increasingly appreciated; however, its true potential is yet to be fully realized. Alongside the growing clinical and research availability of sequencing technologies, precise deep and scalable phenotyping is required to serve unmet need in genetic and rare diseases. To improve the lives of individuals affected with rare diseases through deep phenotyping, global big data interrogation is necessary to aid our understanding of disease biology, assist diagnosis, and develop targeted treatment strategies. This includes the application of cutting-edge machine learning methods to image data. As with most digital tools employed in health care, there are ethical and data governance challenges associated with using identifiable personal image data. There are also risks with failing to deliver on the patient benefits of these new technologies, the biggest of which is posed by data siloing. The Minerva Initiative has been designed to enable the public good of deep phenotyping while mitigating these ethical risks. Its open structure, enabling collaboration and data sharing between individuals, clinicians, researchers and private enterprise, is key for delivering precision public health.

Differentiation of MISSLA and Fanconi anaemia by computer-aided image analysis and presentation of two novel MISSLA siblings

Variants in DONSON were recently identified as the cause of microcephaly, short stature, and limb abnormalities syndrome (MISSLA). The clinical spectra of MISSLA and Fanconi anaemia (FA) strongly overlap. For that reason, some MISSLA patients have been clinically diagnosed with FA. Here, we present the clinical data of siblings with MISSLA featuring a novel DONSON variant and summarize the current literature on MISSLA. Additionally, we perform computer-aided image analysis using the DeepGestalt technology to test how distinct the facial features of MISSLA and FA patients are. We show that MISSLA has a specific facial gestalt. Notably, we find that also FA patients feature facial characteristics recognizable by computer-aided image analysis. We conclude that computer-assisted image analysis improves diagnostic precision in both MISSLA and FA.

Molecular and phenotypic spectrum of Noonan syndrome in Chinese patients

Noonan syndrome (NS) is a common autosomal dominant/recessive disorder. No large-scale study has been conducted on NS in China, which is the most populous country in the world. Next-generation sequencing (NGS) was used to identify pathogenic variants in patients that exhibited NS-related phenotypes. We assessed the facial features and clinical manifestations of patients with pathogenic or likely pathogenic variants in the RAS-MAPK signaling pathway. Gene-related Chinese NS facial features were described using artificial intelligence (AI).NGS identified pathogenic variants in 103 Chinese patients in eight NS-related genes: PTPN11 (48.5%), SOS1 (12.6%), SHOC2 (11.7%), KRAS (9.71%), RAF1 (7.77%), RIT1 (6.8%), CBL (0.97%), NRAS (0.97%), and LZTR1 (0.97%). Gene-related facial representations showed that each gene was associated with different facial details. Eight novel pathogenic variants were detected and clinical features because of specific genetic variants were reported, including hearing loss, cancer risk due to a PTPN11 pathogenic variant, and ubiquitous abnormal intracranial structure due to SHOC2 pathogenic variants. NGS facilitates the diagnosis of NS, especially for patients with mild/moderate and atypical symptoms. Our study describes the genotypic and phenotypic spectra of NS in China, providing new insights into distinctive clinical features due to specific pathogenic variants.

PEDIA: prioritization of exome data by image analysis

Purpose

Phenotype information is crucial for the interpretation of genomic variants. So far it has only been accessible for bioinformatics workflows after encoding into clinical terms by expert dysmorphologists.

Methods

Here, we introduce an approach driven by artificial intelligence that uses portrait photographs for the interpretation of clinical exome data. We measured the value added by computer-assisted image analysis to the diagnostic yield on a cohort consisting of 679 individuals with 105 different monogenic disorders. For each case in the cohort we compiled frontal photos, clinical features, and the disease-causing variants, and simulated multiple exomes of different ethnic backgrounds.

Results

The additional use of similarity scores from computer-assisted analysis of frontal photos improved the top 1 accuracy rate by more than 20–89% and the top 10 accuracy rate by more than 5–99% for the disease-causing gene.

Conclusion

Image analysis by deep-learning algorithms can be used to quantify the phenotypic similarity (PP4 criterion of the American College of Medical Genetics and Genomics guidelines) and to advance the performance of bioinformatics pipelines for exome analysis.

Evaluation of Face2Gene using facial images of patients with congenital dysmorphic syndromes recruited in Japan

An increasing number of genetic syndromes present a challenge to clinical geneticists. A deep learning-based diagnosis assistance system, Face2Gene, utilizes the aggregation of "gestalt," comprising data summarizing features of patients' facial images, to suggest candidate syndromes. Because Face2Gene's results may be affected by ethnicity and age at which training facial images were taken, the system performance for patients in Japan is still unclear. Here, we present an evaluation of Face2Gene using the following two patient groups recruited in Japan: Group 1 consisting of 74 patients with 47 congenital dysmorphic syndromes, and Group 2 consisting of 34 patients with Down syndrome. In Group 1, facial recognition failed for 4 of 74 patients, while 13-21 of 70 patients had a diagnosis for which Face2Gene had not been trained. Omitting these 21 patients, for 85.7% (42/49) of the remainder, the correct syndrome was identified within the top 10 suggested list. In Group 2, for the youngest facial images taken for each of the 34 patients, Down syndrome was successfully identified as the highest-ranking condition using images taken from newborns to those aged 25 years. For the oldest facial images taken at ≥20 years in each of 17 applicable patients, Down syndrome was successfully identified as the highest- and second-highest-ranking condition in 82.2% (14/17) and 100% (17/17) of the patients using images taken from 20 to 40 years. These results suggest that Face2Gene in its current format is already useful in suggesting candidate syndromes to clinical geneticists, using patients with congenital dysmorphic syndromes in Japan.

The facial dysmorphology analysis technology in intellectual disability syndromes related to defects in the histones modifiers

Genetic syndromes are frequently associated with Intellectual Disability (ID), as well as craniofacial dysmorphisms. A group of ID syndromes with typical abnormal face related to chromatin remodeling defects, have been recognized, coining the term chromatinopathies. This is a molecular heterogeneous subset of congenital disorders caused by mutations of the various components of the Chromatin-Marking System (CMS), including modifiers of DNA and chromatin remodelers. We performed a phenotypic study on a sample of 120 individuals harboring variants in genes codifying for the histones enzymes, using the DeepGestalt technology. Three experiments (two multiclass comparison experiments and a frontal face-crop analysis) were conducted, analyzing respectively a total of 181 pediatric images in the first comparison experiment and 180 in the second, all individuals belonging predominantly to Caucasian population. The classification results were expressed in terms of the area under the curve (AUC) of the receiver-operating-characteristic curve (ROC). Significant values of AUC and low p-values were registered for all syndromes in the three experiments, in comparison with each other, with other ID syndromes characterized by recognizable craniofacial dysmorphisms and with unaffected controls. Final findings indicated that this group of diseases is characterized by distinctive dysmorphisms, which result pathognomonic. A correct interrogation and use of adequate informatics aids, could become a valid support for clinicians.

The utility of phenomics in diagnosis of inherited metabolic disorders

Inherited metabolic disorders (IMDs) are debilitating inherited diseases, with phenotypic, biochemical and genetic heterogeneity, frequently leading to prolonged diagnostic odysseys. Mitochondrial disorders represent one of the most severe classes of IMDs, wherein defects in >350 genes lead to multi-system disease. Diagnostic rates have improved considerably following the adoption of next-generation sequencing (NGS) technologies, but are still far from perfect. Phenomic annotation is an emerging concept which is being utilised to enhance interpretation of NGS results. To test whether phenomic correlations have utility in mitochondrial disease and IMDs, we created a gene-to-phenotype interaction network with searchable elements, for Leigh syndrome, a frequently observed paediatric mitochondrial disorder. The Leigh Map comprises data on 92 genes and 275 phenotypes standardised in human phenotype ontology terms, with 80% predictive accuracy. This commentary highlights the usefulness of the Leigh Map and similar resources and the challenges associated with integrating phenomic technologies into clinical practice.

From gestalt to gene: early predictive dysmorphic features of PMM2-CDG

Introduction

Phosphomannomutase-2 deficiency (PMM2-CDG) is associated with a recognisable facial pattern. There are no early severity predictors for this disorder and no phenotype–genotype correlation. We performed a detailed dysmorphology evaluation to describe facial gestalt and its changes over time, to train digital recognition facial analysis tools and to identify early severity predictors.

Methods

Paediatric PMM2-CDG patients were evaluated and compared with controls. A computer-assisted recognition tool was trained. Through the evaluation of dysmorphic features (DFs), a simple categorisation was created and correlated with clinical and neurological scores, and neuroimaging.

Results

Dysmorphology analysis of 31 patients (4–19 years of age) identified eight major DFs (strabismus, upslanted eyes, long fingers, lipodystrophy, wide mouth, inverted nipples, long philtrum and joint laxity) with predictive value using receiver operating characteristic (ROC) curveanalysis (p<0.001). Dysmorphology categorisation using lipodystrophy and inverted nipples was employed to divide patients into three groups that are correlated with global clinical and neurological scores, and neuroimaging (p=0.005, 0.003 and 0.002, respectively). After Face2Gene training, PMM2-CDG patients were correctly identified at different ages.

Conclusions

PMM2-CDG patients’ DFs are consistent and inform about clinical severity when no clear phenotype–genotype correlation is known. We propose a classification of DFs into major and minor with diagnostic risk implications. At present, Face2Gene is useful to suggest PMM2-CDG. Regarding the prognostic value of DFs, we elaborated a simple severity dysmorphology categorisation with predictive value, and we identified five major DFs associated with clinical severity. Both dysmorphology and digital analysis may help physicians to diagnose PMM2-CDG sooner.

Characterization of glycosylphosphatidylinositol biosynthesis defects by clinical features, flow cytometry, and automated image analysis

Background

Glycosylphosphatidylinositol biosynthesis defects (GPIBDs) cause a group of phenotypically overlapping recessive syndromes with intellectual disability, for which pathogenic mutations have been described in 16 genes of the corresponding molecular pathway. An elevated serum activity of alkaline phosphatase (AP), a GPI-linked enzyme, has been used to assign GPIBDs to the phenotypic series of hyperphosphatasia with mental retardation syndrome (HPMRS) and to distinguish them from another subset of GPIBDs, termed multiple congenital anomalies hypotonia seizures syndrome (MCAHS). However, the increasing number of individuals with a GPIBD shows that hyperphosphatasia is a variable feature that is not ideal for a clinical classification.

Methods

We studied the discriminatory power of multiple GPI-linked substrates that were assessed by flow cytometry in blood cells and fibroblasts of 39 and 14 individuals with a GPIBD, respectively. On the phenotypic level, we evaluated the frequency of occurrence of clinical symptoms and analyzed the performance of computer-assisted image analysis of the facial gestalt in 91 individuals.

Results

We found that certain malformations such as Morbus Hirschsprung and diaphragmatic defects are more likely to be associated with particular gene defects (PIGV, PGAP3, PIGN). However, especially at the severe end of the clinical spectrum of HPMRS, there is a high phenotypic overlap with MCAHS. Elevation of AP has also been documented in some of the individuals with MCAHS, namely those with PIGA mutations. Although the impairment of GPI-linked substrates is supposed to play the key role in the pathophysiology of GPIBDs, we could not observe gene-specific profiles for flow cytometric markers or a correlation between their cell surface levels and the severity of the phenotype. In contrast, it was facial recognition software that achieved the highest accuracy in predicting the disease-causing gene in a GPIBD.

Conclusions

Due to the overlapping clinical spectrum of both HPMRS and MCAHS in the majority of affected individuals, the elevation of AP and the reduced surface levels of GPI-linked markers in both groups, a common classification as GPIBDs is recommended. The effectiveness of computer-assisted gestalt analysis for the correct gene inference in a GPIBD and probably beyond is remarkable and illustrates how the information contained in human faces is pivotal in the delineation of genetic entities.

Electronic supplementary material

The online version of this article (doi:10.1186/s13073-017-0510-5) contains supplementary material, which is available to authorized users.