Machine learning approaches to the identification of children affected by prenatal alcohol exposure: A narrative review.

Fetal alcohol spectrum disorders (FASDs) affect at least 0.8% of the population globally. The diagnosis of FASD is uniquely complex, with a heterogeneous physical and neurobehavioral presentation that requires multidisciplinary expertise for diagnosis. Many researchers have begun to incorporate machine learning approaches into FASD research to identify children who are affected by prenatal alcohol exposure, including those with FASD. This narrative review highlights these efforts. Following an introduction to machine learning, we summarize examples from the literature of neurobehavioral screening tools and physiologic markers of exposure. We discuss individual efforts, including models that classify FASD based on parent-reported neurocognitive or behavioral questionnaires, 3D facial imaging, brain imaging, DNA methylation patterns, microRNA profiles, cardiac orienting response, and dysmorphic facial features. We highlight model performance and discuss the limitations of these approaches. We conclude by considering the scalability of these approaches and how these machine learning models, largely developed from clinical samples or highly exposed birth cohorts, may perform in the general population.

Secondary physical features in children with FASD.

OBJECTIVE: The diagnoses included within the umbrella term fetal alcohol spectrum disorders (FASD), are based on the documentation of prenatal alcohol exposure (PAE), growth deficits and a pattern of dysmorphic physical features and neurobehavioral impairments. Although 3 key facial features (short palpebral fissures, a smooth philtrum and a thin vermilion of the upper lip) are the only dysmorphic features taken into account for the diagnosis of Fetal Alcohol Syndrome (FAS) or partial FAS (pFAS), several other features are commonly seen in individuals with these diagnoses. The goals of our study were to determine if some of these secondary physical features also occur more frequently in children with alcohol-related neurodevelopmental disorder (ARND) relative to controls, and if a cluster of these features combined in a dysmorphology score could be used to identify those negatively impacted by PAE but who do not have the cardinal physical features that led to a diagnosis of FAS or pFAS. METHODS: Among 2681 children recruited for the Collaboration on Fetal Alcohol Spectrum Disorders Prevalence (CoFASP) study, 1726 had an FASD or sufficient evidence of PAE having occurred or not in their pregnancy. Children were then categorized into groups using the modified Hoyme diagnostic criteria (FAS (n = 24), pFAS (n = 99) and ARND (n = 87), and No FASD (n = 1516), including those with No FASD and a history of PAE (No FASD/PAE, n = 498) and those with No FASD and no history of PAE (No FASD/No PAE, n = 1018). The frequencies of 26 secondary dysmorphic features were compared among these groups, both individually and combined in non-weighted and weighted dysmorphic scores. Correlations of the total dysmorphic scores with an index of overall cognitive ability were also compared by group status. RESULTS: Several of these features were significantly more frequent in children with FAS than in those with No FASD diagnosis with or without PAE but not in comparison to those with ARND. The number of features was also significantly higher in the FAS group as compared to all other groups for both weighted and unweighted dysmorphology scores but were not higher in the group with ARND when compared to the groups with No FASD either in the presence or absence of PAE. Although not diagnostic, higher total dysmorphology scores were predictive of lower general cognitive abilities in the group with ARND, suggesting severity of alcohol-related dysmorphology is predictive of severity of alcohol-related neurobehavioral impairment. CONCLUSION: Secondary physical features were not more frequent in children with ARND compared to children without an FASD diagnosis but were a marker for lower cognitive function. The use of secondary physical features to support a diagnosis of ARND was not supported in this sample.

Mutation in the Bone Morphogenetic Protein signaling pathway sensitize zebrafish and humans to ethanol-induced jaw malformations.

Fetal Alcohol Spectrum Disorders (FASD) describe ethanol-induced developmental defects including craniofacial malformations. While ethanol-sensitive genetic mutations contribute to facial malformations, the impacted cellular mechanisms remain unknown. Bmp signaling is a key regulator of epithelial morphogenesis driving facial development, providing a possible ethanol-sensitive mechanism. We found that zebrafish mutants for Bmp signaling components are ethanol-sensitive and affect anterior pharyngeal endoderm shape and gene expression, indicating ethanol-induced malformations of the anterior pharyngeal endoderm cause facial malformations. Integrating FASD patient data, we provide the first evidence that variants in the human Bmp receptor gene BMPR1B associate with ethanol-related differences in jaw volume. Our results show that ethanol exposure disrupts proper morphogenesis of, and tissue interactions between, facial epithelia that mirror overall viscerocranial shape changes and are predictive for Bmp-ethanol associations in human jaw development. Our data provide a mechanistic paradigm linking ethanol to disrupted epithelial cell behaviors that underlie facial defects in FASD.

Facial deformation following treatment for pediatric head and neck rhabdomyosarcoma; the difference between treatment modalities. Results of a trans-Atlantic, multicenter cross-sectional cohort study.

BACKGROUND: The four different local therapy strategies used for head and neck rhabdomyosarcoma (HNRMS) include proton therapy (PT), photon therapy (RT), surgery with radiotherapy (Paris-method), and surgery with brachytherapy (AMORE). Local control and survival is comparable; however, the impact of these different treatments on facial deformation is still poorly understood. This study aims to quantify facial deformation and investigates the differences in facial deformation between treatment modalities. METHODS: Across four European and North American institutions, HNRMS survivors treated between 1990 and 2017, more than 2 years post treatment, had a 3D photograph taken. Using dense surface modeling, we computed facial signatures for each survivor to show facial deformation relative to 35 age-sex-ethnicity-matched controls. Additionally, we computed individual facial asymmetry. FINDINGS: A total of 173 HNRMS survivors were included, survivors showed significantly reduced facial growth (p < .001) compared to healthy controls. Partitioned by tumor site, there was reduced facial growth in survivors with nonparameningeal primaries (p = .002), and parameningeal primaries (p ≤.001), but not for orbital primaries (p = .080) All patients were significantly more asymmetric than healthy controls, independent of treatment modality (p ≤ .001). There was significantly more facial deformation in orbital patients when comparing RT to AMORE (p = .046). In survivors with a parameningeal tumor, there was significantly less facial deformation in PT when compared to RT (p = .009) and Paris-method (p = .007). INTERPRETATION: When selecting optimal treatment, musculoskeletal facial outcomes are an expected difference between treatment options. These anticipated differences are currently based on clinicians' bias, expertise, and experience. These data supplement clinician judgment with an objective analysis highlighting the impact of patient age and tumor site between existing treatment options.

Comparison of three systems for the diagnosis of fetal alcohol spectrum disorders in a community sample.

BACKGROUND: It is estimated that 1%-5% of children in the United States are affected by prenatal alcohol exposure while only a small percentage are so identified in clinical practice. One explanation for this discrepancy may be the way in which diagnostic criteria are operationalized. METHODS: To evaluate the extent to which three commonly used systems for the diagnosis of Fetal Alcohol Spectrum Disorder (FASD) consistently identified children in a community sample, data from the Collaboration on Fetal Alcohol Spectrum Disorders Prevalence (COFASP) study were re-analyzed. In the data set, there were 2325 children with variables necessary to allow diagnosis by three systems commonly used in North America. These systems were (1) that used by COFASP, which is a revised modification of the Institute of Medicine's recommendations, (2) the 4-Digit Code, and (3) the most recent Canadian Guidelines. To determine the degree of association among these classifications, the Fleiss Multirater Kappa measure of agreement was applied. RESULTS: Among these three systems, 408 children were classified as FASD, 208 by the CoFASP system, 319 by the 4-Digit Code, and 28 by the Canadian Guidelines. Agreement among the findings from the three systems varied from slight to fair. CONCLUSIONS: These results indicate a lack of consistency in these approaches to FASD diagnosis. Discrepancies result from differences in specifying the criteria used to define the diagnosis, including growth, physical features, neurobehavior, and alcohol-use thresholds. The question of their relative accuracy cannot be resolved without reference to a measure of validity that does not currently exist, and this suggests the need for a more empirically based diagnostic schema.

Facial Anatomical Landmark Detection Using Regularized Transfer Learning With Application to Fetal Alcohol Syndrome Recognition.

Fetal alcohol syndrome (FAS) caused by prenatal alcohol exposure can result in a series of cranio-facial anomalies, and behavioral and neurocognitive problems. Current diagnosis of FAS is typically done by identifying a set of facial characteristics, which are often obtained by manual examination. Anatomical landmark detection, which provides rich geometric information, is important to detect the presence of FAS associated facial anomalies. This imaging application is characterized by large variations in data appearance and limited availability of labeled data. Current deep learning-based heatmap regression methods designed for facial landmark detection in natural images assume availability of large datasets and are therefore not well-suited for this application. To address this restriction, we develop a new regularized transfer learning approach that exploits the knowledge of a network learned on large facial recognition datasets. In contrast to standard transfer learning which focuses on adjusting the pre-trained weights, the proposed learning approach regularizes the model behavior. It explicitly reuses the rich visual semantics of a domain-similar source model on the target task data as an additional supervisory signal for regularizing landmark detection optimization. Specifically, we develop four regularization constraints for the proposed transfer learning, including constraining the feature outputs from classification and intermediate layers, as well as matching activation attention maps in both spatial and channel levels. Experimental evaluation on a collected clinical imaging dataset demonstrate that the proposed approach can effectively improve model generalizability under limited training samples, and is advantageous to other approaches in the literature.

Increased facial asymmetry in focal epilepsies associated with unilateral lesions.

The epilepsies are now conceptualized as network disruptions: focal epilepsies are considered to have network alterations in the hemisphere of seizure onset, whilst generalized epilepsies are considered to have bi-hemispheric network changes. Increasingly, many epilepsies are also considered to be neurodevelopmental disorders, with early changes in the brain underpinning seizure biology. The development of the structure of the face is influenced by complex molecular interactions between surface ectoderm and underlying developing forebrain and neural crest cells. This influence is likely to continue postnatally, given the evidence of facial growth changes over time in humans until at least 18 years of age. In this case-control study, we hypothesized that people with lateralized focal epilepsies (i.e. unilateral network changes) have an increased degree of facial asymmetry, compared with people with generalized epilepsies or controls without epilepsy. We applied three-dimensional stereophotogrammetry and dense surface models to evaluate facial asymmetry in people with epilepsy, aiming to generate new tools to explore pathophysiological mechanisms in epilepsy. We analysed neuroimaging data to explore the correlation between face and brain asymmetry. We consecutively recruited 859 people with epilepsy attending the epilepsy clinics at a tertiary referral centre. We used dense surface modelling of the full face and signature analyses of three-dimensional facial photographs to analyse facial differences between 378 cases and 205 healthy controls. Neuroimaging around the time of the facial photograph was available for 234 cases. We computed the brain asymmetry index between contralateral regions. Cases with focal symptomatic epilepsy associated with unilateral lesions showed greater facial asymmetry compared to controls (P = 0.0001, two-sample t-test). This finding was confirmed by linear regression analysis after controlling for age and gender. We also found a significant correlation between duration of illness and the brain asymmetry index of total average cortical thickness (r = -0.19, P = 0.0075) but not for total average surface area (r = 0.06, P = 0.3968). There was no significant correlation between facial asymmetry and asymmetry of regional cortical thickness or surface area. We propose that the greater facial asymmetry in cases with focal epilepsy caused by unilateral abnormality might be explained by early unilateral network disruption, and that this is independent of underlying brain asymmetry. Three-dimensional stereophotogrammetry and dense surface modelling are a novel powerful phenotyping tool in epilepsy that may permit greater understanding of pathophysiology in epilepsy, and generate further insights into the development of cerebral networks underlying epilepsy, and the genetics of facial and neural development.

3D analysis of facial morphology in Dutch children with cancer.

Background and Objective; Genetic risk factors for childhood cancer may also influence facial morphology. 3D photography can be used in the recognition of differences in face shape among individuals. In previous research, 3D facial photography was used to identify increased facial asymmetry and greater deviation from normal facial morphology in a group of individuals with distinct morphological features who had childhood cancer compared to healthy controls. In this study, we aim to determine whether there is a difference in facial morphology between children with cancer without previously selected morphological features and healthy controls, detected with 3D facial photography. METHODS: Facial 3D photographic images were obtained of children with a newly diagnosed malignancy. The resulting sample comprised 13 different cancer types. Patients were excluded if they had a known genetic cause of the cancer. Patients were compared to healthy controls, matched for sex, age and ethnic background. The degree of asymmetry and overall deviation of an individual's face from an age and sex typical control face were measured. RESULTS: A total of 163 patients of European descent were included. No significant difference in asymmetry between patients and controls could be identified. On average, patients deviated more from an age and sex typical face than the controls. CONCLUSION: This study shows that children with cancer deviate more than controls, possibly suggesting a higher prevalence of genetic anomalies within this group. The results suggest that this is not sufficient to discriminate patients from controls. Further research is necessary to explore the patterns of individual variation among the overall deviation of patients and controls.

Fetal alcohol spectrum disorders: an overview of current evidence and activities in the UK.

Estimates for the UK suggest that alcohol consumption during pregnancy and prevalence of fetal alcohol spectrum disorder (FASD)-the most common neurodevelopmental condition-are high. Considering the significant health and social impacts of FASD, there is a public health imperative to prioritise prevention, interventions and support. In this article, we outline the current state of play regarding FASD knowledge and research in the UK, which is characterised by a lack of evidence, a lack of dedicated funding and services, and consequently little policy formulation and strategic direction. We highlight progress made to date, as well as current knowledge and service gaps to propose a way forward for UK research.

Objectifying Micrognathia Using Three-Dimensional Photogrammetric Analysis.

BACKGROUND: Micrognathia occurs isolated and as part of entities like Robin sequence (RS). An objective measurement of mandible size and growth is needed to determine the degree of micrognathia and enable a comparison of treatment outcomes. A pilot study was conducted to investigate the usability of 3-dimensional (3D) facial photogrammetry, a fast, noninvasive method, to estimate mandible size and growth in a small cohort of newborns and infants. METHODS: Exterior mandibular volume was estimated using a tetrahedron defined by 4 facial landmarks. Twelve patients with RS with different etiologies were selected and photogrammetric images were obtained prospectively in 3 patients with RS in whom mandibular growth in the first year of life was determined. We used 3 tetrahedra defined by 6 landmarks on mandibular computed tomography (CT) scans to estimate an interior mandibular volume, which we compared to the exterior mandibular volume in 10 patients. RESULTS: The exterior mandibular volume using 3D photography could be determined in all patients. Signature heat maps allowed visualization of facial dysmorphism in 3D; signature graphs demonstrated similarities of facial dysmorphism in patients with the same etiology and differences from those with other diagnoses and from controls. The correlation between interior (3D photogrammetry) and exterior mandibular volumes (CT imaging) was 0.8789. CONCLUSION: The 3D facial photogrammetry delineates the general facial characteristics in patients with different syndromes involving micrognathia, and can objectively estimate mandibular volume and growth, with excellent correlation with bony measurement. It has been concluded that 3D facial photogrammetry could be a clinically effective instrument for delineating and quantifying micrognathia.

Combined Face-Brain Morphology and Associated Neurocognitive Correlates in Fetal Alcohol Spectrum Disorders.

BACKGROUND: Since the 1970s, a range of facial, neurostructural, and neurocognitive adverse effects have been shown to be associated with prenatal alcohol exposure. Typically, these effects are studied individually and not in combination. Our objective is to improve the understanding of the teratogenic effects of prenatal alcohol exposure by simultaneously considering face-brain morphology and neurocognitive measures. METHODS: Participants were categorized as control (n = 47), fetal alcohol syndrome (FAS, n = 22), or heavily exposed (HE) prenatally, but not eligible for a FAS diagnosis (HE, n = 50). Structural brain MRI images and high-resolution 3D facial images were analyzed using dense surface models of features of the face and surface shape of the corpus callosum (CC) and caudate nucleus (CN). Asymmetry of the CN was evaluated for correlations with neurocognitive measures. RESULTS: (i) Facial growth delineations for FAS, HE, and controls are replicated for the CN and the CC. (ii) Concordance of clinical diagnosis and face-based control-FAS discrimination improves when the latter is combined with specific brain regions. In particular, midline facial regions discriminate better when combined with a midsagittal profile of the CC. (iii) A subset of HE individuals was identified with FAS-like CN dysmorphism. The average of this HE subset was FAS-like in its facial dysmorphism. (iv) Right-left asymmetry found in the CNs of controls is not apparent for FAS, is diminished for HE, and correlates with neurocognitive measures in the combined FAS and HE population. CONCLUSIONS: Shape analysis which combines facial regions with the CN, and with the CC, better identify those with FAS. CN asymmetry was reduced for FAS compared to controls and is strongly associated with general cognitive ability, verbal learning, and recall in those with prenatal alcohol exposure. This study further extends the brain-behavior relationships known to be vulnerable to alcohol teratogenesis.

L1 coupling to ankyrin and the spectrin-actin cytoskeleton modulates ethanol inhibition of L1 adhesion and ethanol teratogenesis.

Ethanol causes fetal alcohol spectrum disorders (FASDs) partly by inhibiting cell adhesion mediated by the L1 neural cell adhesion molecule. Ethanol interacts with an alcohol binding pocket in the L1 extracellular domain (ECD), and dephosphorylation of S1248 in the L1 cytoplasmic domain (CD) renders L1 adhesion insensitive to inhibition by ethanol (L1 insensitive). The mechanism underlying this inside-out signaling is unknown. Here we show that phosphorylation of the human L1-CD at S1152, Y1176, S1181, and S1248 renders L1 sensitive to ethanol by promoting L1 coupling with ankyrin-G and the spectrin-actin cytoskeleton. Knockdown of ankyrin-G or L1 mutations that uncouple L1 from ankyrin reduce L1 sensitivity to ethanol, but not methanol, consistent with a small conformational change in the extracellular alcohol binding pocket. Phosphorylation of Y1176 and ankyrin-G coupling with L1 are higher in NIH/3T3 clonal cell lines in which ethanol inhibits L1 adhesion than in ethanol-resistant NIH/3T3 clonal cell lines. Similarly, phosphorylation of Y1176 is higher in C57BL/6J mice that are sensitive to ethanol teratogenesis than in ethanol resistant C57BL/6N mice. Finally, polymorphisms in genes that encode ankyrin-G and p90rsk, a kinase that phosphorylates S1152, are linked to facial dysmorphology in children with heavy prenatal ethanol exposure. These findings indicate that genes that regulate L1 coupling to ankyrin may influence susceptibility to FASD.-Dou, X., Menkari, C., Mitsuyama, R., Foroud, T., Wetherill, L., Hammond, P., Suttie, M., Chen, X., Chen, S.-Y., Charness, M. E., Collaborative Initiative on Fetal Alcohol Spectrum Disorders. L1 coupling to ankyrin and the spectrin-actin cytoskeleton modulates ethanol inhibition of L1 adhesion and ethanol teratogenesis.

Facial Curvature Detects and Explicates Ethnic Differences in Effects of Prenatal Alcohol Exposure.

BACKGROUND: Our objective is to help clinicians detect the facial effects of prenatal alcohol exposure by developing computer-based tools for screening facial form. METHODS: All 415 individuals considered were evaluated by expert dysmorphologists and categorized as (i) healthy control (HC), (ii) fetal alcohol syndrome (FAS), or (iii) heavily prenatally alcohol exposed (HE) but not clinically diagnosable as FAS; 3D facial photographs were used to build models of facial form to support discrimination studies. Surface curvature-based delineations of facial form were introduced. RESULTS: (i) Facial growth in FAS, HE, and control subgroups is similar in both cohorts. (ii) Cohort consistency of agreement between clinical diagnosis and HC-FAS facial form classification is lower for midline facial regions and higher for nonmidline regions. (iii) Specific HC-FAS differences within and between the cohorts include: for HC, a smoother philtrum in Cape Coloured individuals; for FAS, a smoother philtrum in Caucasians; for control-FAS philtrum difference, greater homogeneity in Caucasians; for control-FAS face difference, greater homogeneity in Cape Coloured individuals. (iv) Curvature changes in facial profile induced by prenatal alcohol exposure are more homogeneous and greater in Cape Coloureds than in Caucasians. (v) The Caucasian HE subset divides into clusters with control-like and FAS-like facial dysmorphism. The Cape Coloured HE subset is similarly divided for nonmidline facial regions but not clearly for midline structures. (vi) The Cape Coloured HE subset with control-like facial dysmorphism shows orbital hypertelorism. CONCLUSIONS: Facial curvature assists the recognition of the effects of prenatal alcohol exposure and helps explain why different facial regions result in inconsistent control-FAS discrimination rates in disparate ethnic groups. Heavy prenatal alcohol exposure can give rise to orbital hypertelorism, supporting a long-standing suggestion that prenatal alcohol exposure at a particular time causes increased separation of the brain hemispheres with a concomitant increase in orbital separation.

Facial asymmetry in head and neck rhabdomyosarcoma survivors.

INTRODUCTION: Radiotherapy is essential for achieving and maintaining local control in head and neck rhabdomyosarcoma (HNRMS) patients. However, radiotherapy may cause outgrowth disturbances of facial bone and soft tissue, resulting in facial asymmetry. The aim of this study was to develop a method to visualize and measure facial asymmetry in HNRMS survivors using three-dimensional (3D) imaging techniques. METHODS: Facial deformity was evaluated in a multidisciplinary clinical assessment of 75 HNRMS survivors, treated with external beam radiotherapy (EBRT, n = 26) or Ablative surgery, MOulage brachytherapy, and REconstruction (AMORE, n = 49). Individual facial asymmetry was measured using 3D photogrammetry and expressed in a raw asymmetry index and a normalized sex-age-ethnicity-matched asymmetry signature weight. Facial asymmetry was also compared between British and Dutch controls and between survivors and their matched controls. RESULTS: Facial asymmetry was more pronounced with increasing age (P < 0.01) in British controls compared with Dutch controls (P = 0.04). Survivors developed more facial asymmetry than matched controls (P < 0.001). The clinical assessment of facial deformity correlated with the raw asymmetry index (r = 0.60, P < 0.001). DISCUSSION: 3D imaging can be used for objective measurement of facial asymmetry in HNRMS survivors. The raw asymmetry index correlated with a clinical assessment of facial deformity. Comparisons between treatment groups seemed inappropriate given the differences in facial asymmetry between British and Dutch controls. In future studies, pretreatment images could act as matched controls for posttreatment evaluation.

Combined soft and skeletal tissue modelling of normal and dysmorphic midface postnatal development.

BACKGROUND: Midface hypoplasia as exemplified by Treacher Collins Syndrome (TCS) can impair appearance and function. Reconstruction involves multiple invasive surgeries with variable long-term outcomes. This study aims to describe normal and dysmorphic midface postnatal development through combined modelling of skeletal and soft tissues and to develop a surgical evaluation tool. MATERIALS AND METHODS: Midface skeletal and soft tissue surfaces were extracted from computed tomography scans of 52 control and 14 TCS children, then analysed using dense surface modelling. The model was used to describe midface growth, morphology, and asymmetry, then evaluate postoperative outcomes. RESULTS: Parameters responsible for the greatest variation in midface size and shape showed differences between TCS and controls with close alignment between skeletal and soft tissue models. TCS children exhibited midface dysmorphology and hypoplasia when compared with controls. Asymmetry was also significantly higher in TCS midfaces. Combined modelling was used to evaluate the impact of surgery in one TCS individual who showed normalisation immediately after surgery but reversion towards TCS dysmorphology after 1 year. CONCLUSION: This is the first quantitative analysis of postnatal midface development using combined modelling of skeletal and soft tissues. We also provide an approach for evaluation of surgical outcomes, laying the foundations for future development of a preoperative planning tool.

3D morphometry aids facial analysis of individuals with a childhood cancer.

A group of patients who had cancer as a child were previously found to have distinct patterns of morphological abnormalities. In this study, we investigated the added value of 3D shape analysis to characterize their facial morphology. Primarily, we showed in an objective and quantitative manner that the overall facial dysmorphism of the individuals who had had a childhood cancer was significantly greater than that of the controls. We also demonstrated how the same approach can be used to detect a similar disparity for a more localized malar region comprising customized disconnected patches defined on both sides of the face. In addition, by comparing original face surfaces to their mirrored forms, we confirmed that the patient group had significantly greater facial asymmetry than the controls. Each of these results made use of surface shape differences not detectable by simple linear or angular characteristics as might be used in analyses based on measures captured manually or derived from landmarks annotating 2D photographic images. We conclude that 3D morphometric analysis of a relatively small heterogeneous patient group can further delineate face shape differences from typically developing individuals that are too subtle or geometrically complex to identify or quantify objectively with conventional clinical and anthropometric approaches. © 2016 Wiley Periodicals, Inc.

Opposite effects on facial morphology due to gene dosage sensitivity.

Sequencing technology is increasingly demonstrating the impact of genomic copy number variation (CNV) on phenotypes. Opposing variation in growth, head size, cognition and behaviour is known to result from deletions and reciprocal duplications of some genomic regions. We propose normative inversion of face shape, opposing difference from a matched norm, as a basis for investigating the effects of gene dosage on craniofacial development. We use dense surface modelling techniques to match any face (or part of a face) to a facial norm of unaffected individuals of matched age, sex and ethnicity and then we reverse the individual's face shape differences from the matched norm to produce the normative inversion. We demonstrate for five genomic regions, 4p16.3, 7q11.23, 11p15, 16p13.3 and 17p11.2, that such inversion for individuals with a duplication or (epi)-mutation produces facial forms remarkably similar to those associated with a deletion or opposite (epi-)mutation of the same region, and vice versa. The ability to visualise and quantify face shape effects of gene dosage is of major benefit for determining whether a CNV is the cause of the phenotype of an individual and for predicting reciprocal consequences. It enables face shape to be used as a relatively simple and inexpensive functional analysis of the gene(s) involved.

Face shape differs in phylogenetically related populations.

3D analysis of facial morphology has delineated facial phenotypes in many medical conditions and detected fine grained differences between typical and atypical patients to inform genotype-phenotype studies. Next-generation sequencing techniques have enabled extremely detailed genotype-phenotype correlative analysis. Such comparisons typically employ control groups matched for age, sex and ethnicity and the distinction between ethnic categories in genotype-phenotype studies has been widely debated. The phylogenetic tree based on genetic polymorphism studies divides the world population into nine subpopulations. Here we show statistically significant face shape differences between two European Caucasian populations of close phylogenetic and geographic proximity from the UK and The Netherlands. The average face shape differences between the Dutch and UK cohorts were visualised in dynamic morphs and signature heat maps, and quantified for their statistical significance using both conventional anthropometry and state of the art dense surface modelling techniques. Our results demonstrate significant differences between Dutch and UK face shape. Other studies have shown that genetic variants influence normal facial variation. Thus, face shape difference between populations could reflect underlying genetic difference. This should be taken into account in genotype-phenotype studies and we recommend that in those studies reference groups be established in the same population as the individuals who form the subject of the study.

Deletions involving genes WHSC1 and LETM1 may be necessary, but are not sufficient to cause Wolf-Hirschhorn Syndrome.

Wolf-Hirschhorn syndrome (WHS) is a complex genetic disorder caused by the loss of genomic material from the short arm of chromosome 4. Genotype-phenotype correlation studies indicated that the loss of genes within 4p16.3 is necessary for expression of the core features of the phenotype. Within this region, haploinsufficiency of the genes WHSC1 and LETM1 is thought to be a major contributor to the pathogenesis of WHS. We present clinical findings for three patients with relatively small (<400 kb) de novo interstitial deletions that overlap WHSC1 and LETM1. 3D facial analysis was performed for two of these patients. Based on our findings, we propose that hemizygosity of WHSC1 and LETM1 is associated with a clinical phenotype characterized by growth deficiency, feeding difficulties, and motor and speech delays. The deletion of additional genes nearby WHSC1 and LETM1 does not result in a marked increase in the severity of clinical features, arguing against their haploinsufficiency. The absence of seizures and typical WHS craniofacial findings in our cohort suggest that deletion of distinct or additional 4p16.3 genes is necessary for expression of these features. Altogether, these results show that although loss-of-function for WHSC1 and/or LETM1 contributes to some of the features of WHS, deletion of additional genes is required for the full expression of the phenotype, providing further support that WHS is a contiguous gene deletion disorder.

Growth hormone, gender and face shape in Prader-Willi syndrome.

Prader-Willi syndrome is a neurodevelopmental disorder resulting from the absence of expression of paternally expressed gene(s) in a highly imprinted region of chromosome 15q11-13. The physical phenotype includes evidence of growth retardation due to relative growth hormone deficiency, small hands and feet, a failure of normal secondary sexual development, and a facial appearance including narrow bifrontal diameter, almond-shaped palpebral fissures, narrow nasal root, and thin upper vermilion with downturned corners of the mouth. Anecdotally, the face of individuals with PWS receiving hGH treatment is said to "normalize." We used dense surface modelling and shape signature techniques to analyze 3D photogrammetric images of the faces of 72 affected and 388 unaffected individuals. We confirmed that adults with Prader-Willi syndrome who had never received human growth supplementation displayed known characteristic facial features. Facial growth was significantly reduced in these adults, especially in males. We demonstrated that following human growth hormone (hGH) supplementation, vertical facial growth of affected individuals falls within the normal range. However, lateral and periorbital face shape and nose shape differences in affected children who have received hGH therapy remain sufficiently strong to be significantly discriminating in comparisons with age-sex matched, unaffected individuals. Finally, we produced evidence that age at initiation and length of treatment with hGH do not appear to play a role in normalization or in consistent alteration of the face shape of affected individuals. This is the first study to provide objective shape analysis of craniofacial effects of hGH therapy in Prader-Willi syndrome.