Mapping genes for human face shape: exploration of univariate phenotyping strategies.

Human facial shape, while strongly heritable, involves both genetic and structural complexity, necessitating precise phenotyping for accurate assessment. Common phenotyping strategies include simplifying 3D facial features into univariate traits such as anthropometric measurements (e.g., inter-landmark distances), unsupervised dimensionality reductions (e.g., principal component analysis (PCA) and auto-encoder (AE) approaches), and assessing resemblance to particular facial gestalts (e.g., syndromic facial archetypes). This study provides a comparative assessment of these strategies in genome-wide association studies (GWASs) of 3D facial shape. Specifically, we investigated inter-landmark distances, PCA and AE-derived latent dimensions, and facial resemblance to random, extreme, and syndromic gestalts within a GWAS of 8,426 individuals of recent European ancestry. Inter-landmark distances exhibit the highest SNP-based heritability as estimated via LD score regression, followed by AE dimensions. Conversely, resemblance scores to extreme and syndromic facial gestalts display the lowest heritability, in line with expectations. Notably, the aggregation of multiple GWASs on facial resemblance to random gestalts reveals the highest number of independent genetic loci. This novel, easy-to-implement phenotyping approach holds significant promise for capturing genetically relevant morphological traits derived from complex biomedical imaging datasets, and its applications extend beyond faces. Nevertheless, these different phenotyping strategies capture different genetic influences on craniofacial shape. Thus, it remains valuable to explore these strategies individually and in combination to gain a more comprehensive understanding of the genetic factors underlying craniofacial shape and related traits. Author Summary: Advancements linking variation in the human genome to phenotypes have rapidly evolved in recent decades and have revealed that most human traits are influenced by genetic variants to at least some degree. While many traits, such as stature, are straightforward to acquire and investigate, the multivariate and multipartite nature of facial shape makes quantification more challenging. In this study, we compared the impact of different facial phenotyping approaches on gene mapping outcomes. Our findings suggest that the choice of facial phenotyping method has an impact on apparent trait heritability and the ability to detect genetic association signals. These results offer valuable insights into the importance of phenotyping in genetic investigations, especially when dealing with highly complex morphological traits.

An interactive atlas of three-dimensional syndromic facial morphology.

Craniofacial phenotyping is critical for both syndrome delineation and diagnosis because craniofacial abnormalities occur in 30% of characterized genetic syndromes. Clinical reports, textbooks, and available software tools typically provide two-dimensional, static images and illustrations of the characteristic phenotypes of genetic syndromes. In this work, we provide an interactive web application that provides three-dimensional, dynamic visualizations for the characteristic craniofacial effects of 95 syndromes. Users can visualize syndrome facial appearance estimates quantified from data and easily compare craniofacial phenotypes of different syndromes. Our application also provides a map of morphological similarity between a target syndrome and other syndromes. Finally, users can upload 3D facial scans of individuals and compare them to our syndrome atlas estimates. In summary, we provide an interactive reference for the craniofacial phenotypes of syndromes that allows for precise, individual-specific comparisons of dysmorphology.

Syndrome-informed phenotyping identifies a polygenic background for achondroplasia-like facial variation in the general population.

Human craniofacial shape is highly variable yet highly heritable with genetic variants interacting through multiple layers of development. Here, we hypothesize that Mendelian phenotypes represent the extremes of a phenotypic spectrum and, using achondroplasia as an example, we introduce a syndrome-informed phenotyping approach to identify genomic loci associated with achondroplasia-like facial variation in the normal population. We compared three-dimensional facial scans from 43 individuals with achondroplasia and 8246 controls to calculate achondroplasia-like facial scores. Multivariate GWAS of the control scores revealed a polygenic basis for normal facial variation along an achondroplasia-specific shape axis, identifying genes primarily involved in skeletal development. Jointly modeling these genes in two independent control samples showed craniofacial effects approximating the characteristic achondroplasia phenotype. These findings suggest that both complex and Mendelian genetic variation act on the same developmentally determined axes of facial variation, providing new insights into the genetic intersection of complex traits and Mendelian disorders.

Precise modulation of transcription factor levels identifies features underlying dosage sensitivity.

Transcriptional regulation exhibits extensive robustness, but human genetics indicates sensitivity to transcription factor (TF) dosage. Reconciling such observations requires quantitative studies of TF dosage effects at trait-relevant ranges, largely lacking so far. TFs play central roles in both normal-range and disease-associated variation in craniofacial morphology; we therefore developed an approach to precisely modulate TF levels in human facial progenitor cells and applied it to SOX9, a TF associated with craniofacial variation and disease (Pierre Robin sequence (PRS)). Most SOX9-dependent regulatory elements (REs) are buffered against small decreases in SOX9 dosage, but REs directly and primarily regulated by SOX9 show heightened sensitivity to SOX9 dosage; these RE responses partially predict gene expression responses. Sensitive REs and genes preferentially affect functional chondrogenesis and PRS-like craniofacial shape variation. We propose that such REs and genes underlie the sensitivity of specific phenotypes to TF dosage, while buffering of other genes leads to robust, nonlinear dosage-to-phenotype relationships.

Comparing 2D and 3D representations for face-based genetic syndrome diagnosis.

Human genetic syndromes are often challenging to diagnose clinically. Facial phenotype is a key diagnostic indicator for hundreds of genetic syndromes and computer-assisted facial phenotyping is a promising approach to assist diagnosis. Most previous approaches to automated face-based syndrome diagnosis have analyzed different datasets of either 2D images or surface mesh-based 3D facial representations, making direct comparisons of performance challenging. In this work, we developed a set of subject-matched 2D and 3D facial representations, which we then analyzed with the aim of comparing the performance of 2D and 3D image-based approaches to computer-assisted syndrome diagnosis. This work represents the most comprehensive subject-matched analyses to date on this topic. In our analyses of 1907 subject faces representing 43 different genetic syndromes, 3D surface-based syndrome classification models significantly outperformed 2D image-based models trained and evaluated on the same subject faces. These results suggest that the clinical adoption of 3D facial scanning technology and continued collection of syndromic 3D facial scan data may substantially improve face-based syndrome diagnosis.

Detecting 3D syndromic faces as outliers using unsupervised normalizing flow models.

Many genetic syndromes are associated with distinctive facial features. Several computer-assisted methods have been proposed that make use of facial features for syndrome diagnosis. Training supervised classifiers, the most common approach for this purpose, requires large, comprehensive, and difficult to collect databases of syndromic facial images. In this work, we use unsupervised, normalizing flow-based manifold and density estimation models trained entirely on unaffected subjects to detect syndromic 3D faces as statistical outliers. Furthermore, we demonstrate a general, user-friendly, gradient-based interpretability mechanism that enables clinicians and patients to understand model inferences. 3D facial surface scans of 2471 unaffected subjects and 1629 syndromic subjects representing 262 different genetic syndromes were used to train and evaluate the models. The flow-based models outperformed unsupervised comparison methods, with the best model achieving an ROC-AUC of 86.3% on a challenging, age and sex diverse data set. In addition to highlighting the viability of outlier-based syndrome screening tools, our methods generalize and extend previously proposed outlier scores for 3D face-based syndrome detection, resulting in improved performance for unsupervised syndrome detection.

A Deep Invertible 3-D Facial Shape Model for Interpretable Genetic Syndrome Diagnosis.

One of the primary difficulties in treating patients with genetic syndromes is diagnosing their condition. Many syndromes are associated with characteristic facial features that can be imaged and utilized by computer-assisted diagnosis systems. In this work, we develop a novel 3D facial surface modeling approach with the objective of maximizing diagnostic model interpretability within a flexible deep learning framework. Therefore, an invertible normalizing flow architecture is introduced to enable both inferential and generative tasks in a unified and efficient manner. The proposed model can be used (1) to infer syndrome diagnosis and other demographic variables given a 3D facial surface scan and (2) to explain model inferences to non-technical users via multiple interpretability mechanisms. The model was trained and evaluated on more than 4700 facial surface scans from subjects with 47 different syndromes. For the challenging task of predicting syndrome diagnosis given a new 3D facial surface scan, age, and sex of a subject, the model achieves a competitive overall top-1 accuracy of 71%, and a mean sensitivity of 43% across all syndrome classes. We believe that invertible models such as the one presented in this work can achieve competitive inferential performance while greatly increasing model interpretability in the domain of medical diagnosis.

Genome-wide analysis of copy number variants and normal facial variation in a large cohort of Bantu Africans.

Similarity in facial characteristics between relatives suggests a strong genetic component underlies facial variation. While there have been numerous studies of the genetics of facial abnormalities and, more recently, single nucleotide polymorphism (SNP) genome-wide association studies (GWASs) of normal facial variation, little is known about the role of genetic structural variation in determining facial shape. In a sample of Bantu African children, we found that only 9% of common copy number variants (CNVs) and 10-kb CNV analysis windows are well tagged by SNPs (r2 ≥ 0.8), indicating that associations with our internally called CNVs were not captured by previous SNP-based GWASs. Here, we present a GWAS and gene set analysis of the relationship between normal facial variation and CNVs in a sample of Bantu African children. We report the top five regions, which had p values ≤ 9.35 × 10-6 and find nominal evidence of independent CNV association (p < 0.05) in three regions previously identified in SNP-based GWASs. The CNV region with strongest association (p = 1.16 × 10-6, 55 losses and seven gains) contains NFATC1, which has been linked to facial morphogenesis and Cherubism, a syndrome involving abnormal lower facial development. Genomic loss in the region is associated with smaller average lower facial depth. Importantly, new loci identified here were not identified in a SNP-based GWAS, suggesting that CNVs are likely involved in determining facial shape variation. Given the plethora of SNP-based GWASs, calling CNVs from existing data may be a relatively inexpensive way to aid in the study of complex traits.

Genome scans of facial features in East Africans and cross-population comparisons reveal novel associations.

Facial morphology is highly variable, both within and among human populations, and a sizable portion of this variation is attributable to genetics. Previous genome scans have revealed more than 100 genetic loci associated with different aspects of normal-range facial variation. Most of these loci have been detected in Europeans, with few studies focusing on other ancestral groups. Consequently, the degree to which facial traits share a common genetic basis across diverse sets of humans remains largely unknown. We therefore investigated the genetic basis of facial morphology in an East African cohort. We applied an open-ended data-driven phenotyping approach to a sample of 2,595 3D facial images collected on Tanzanian children. This approach segments the face into hierarchically arranged, multivariate features that capture the shape variation after adjusting for age, sex, height, weight, facial size and population stratification. Genome scans of these multivariate shape phenotypes revealed significant (p < 2.5 × 10-8) signals at 20 loci, which were enriched for active chromatin elements in human cranial neural crest cells and embryonic craniofacial tissue, consistent with an early developmental origin of the facial variation. Two of these associations were in highly conserved regions showing craniofacial-specific enhancer activity during embryological development (5q31.1 and 12q21.31). Six of the 20 loci surpassed a stricter threshold accounting for multiple phenotypes with study-wide significance (p < 6.25 × 10-10). Cross-population comparisons indicated 10 association signals were shared with Europeans (seven sharing the same associated SNP), and facilitated fine-mapping of causal variants at previously reported loci. Taken together, these results may point to both shared and population-specific components to the genetic architecture of facial variation.

Large-scale open-source three-dimensional growth curves for clinical facial assessment and objective description of facial dysmorphism.

Craniofacial dysmorphism is associated with thousands of genetic and environmental disorders. Delineation of salient facial characteristics can guide clinicians towards a correct clinical diagnosis and understanding the pathogenesis of the disorder. Abnormal facial shape might require craniofacial surgical intervention, with the restoration of normal shape an important surgical outcome. Facial anthropometric growth curves or standards of single inter-landmark measurements have traditionally supported assessments of normal and abnormal facial shape, for both clinical and research applications. However, these fail to capture the full complexity of facial shape. With the increasing availability of 3D photographs, methods of assessment that take advantage of the rich information contained in such images are needed. In this article we derive and present open-source three-dimensional (3D) growth curves of the human face. These are sequences of age and sex-specific expected 3D facial shapes and statistical models of the variation around the expected shape, derived from 5443 3D images. We demonstrate the use of these growth curves for assessing patients and show that they identify normal and abnormal facial morphology independent from age-specific facial features. 3D growth curves can facilitate use of state-of-the-art 3D facial shape assessment by the broader clinical and biomedical research community. This advance in phenotype description will support clinical diagnosis and the understanding of disease pathogenesis including genotype-phenotype relations.

Genome-wide copy number variations in a large cohort of bantu African children.

BACKGROUND: Copy number variations (CNVs) account for a substantial proportion of inter-individual genomic variation. However, a majority of genomic variation studies have focused on single-nucleotide variations (SNVs), with limited genome-wide analysis of CNVs in large cohorts, especially in populations that are under-represented in genetic studies including people of African descent. METHODS: We carried out a genome-wide copy number analysis in > 3400 healthy Bantu Africans from Tanzania. Signal intensity data from high density (> 2.5 million probes) genotyping arrays were used for CNV calling with three algorithms including PennCNV, DNAcopy and VanillaICE. Stringent quality metrics and filtering criteria were applied to obtain high confidence CNVs. RESULTS: We identified over 400,000 CNVs larger than 1 kilobase (kb), for an average of 120 CNVs (SE = 2.57) per individual. We detected 866 large CNVs (≥ 300 kb), some of which overlapped genomic regions previously associated with multiple congenital anomaly syndromes, including Prader-Willi/Angelman syndrome (Type1) and 22q11.2 deletion syndrome. Furthermore, several of the common CNVs seen in our cohort (≥ 5%) overlap genes previously associated with developmental disorders. CONCLUSIONS: These findings may help refine the phenotypic outcomes and penetrance of variations affecting genes and genomic regions previously implicated in diseases. Our study provides one of the largest datasets of CNVs from individuals of African ancestry, enabling improved clinical evaluation and disease association of CNVs observed in research and clinical studies in African populations.

The Genetic Basis of Vitiligo.

Vitiligo is a complex disease in which autoimmune destruction of epidermal melanocytes results in patches of depigmented white skin. Vitiligo has an estimated prevalence of about 0.2-2% in different populations and approximately 0.4% in the European-derived white (EUR) population. The fraction of disease risk attributable to genetic variation, termed heritability, is high, with estimates from family studies in EUR of 0.75-0.83 and from SNP based studies estimated at 0.78. About 70% of genetic risk comes from common genetic variants and about 30% from rare genetic variants. Through candidate gene, genomewide linkage, and genomewide association studies, over 50 vitiligo susceptibility loci have been discovered. These have been combined into a vitiligo polygenic risk score, which has allowed various aspects of vitiligo genetic architecture in the EUR population to be better understood. Vitiligo has thus proved to be a particularly tractable model for investigation of complex disease genetic architecture. Here, we summarize progress to date including dissection of heritability, discovery of vitiligo susceptibility loci through candidate gene, genomewide linkage, and genomewide association studies, relationships to other autoimmune diseases, polygenic architecture of vitiligo risk, vitiligo triggering, and disease onset, and provide suggestions for future directions.

FaceBase 3: analytical tools and FAIR resources for craniofacial and dental research.

The FaceBase Consortium was established by the National Institute of Dental and Craniofacial Research in 2009 as a 'big data' resource for the craniofacial research community. Over the past decade, researchers have deposited hundreds of annotated and curated datasets on both normal and disordered craniofacial development in FaceBase, all freely available to the research community on the FaceBase Hub website. The Hub has developed numerous visualization and analysis tools designed to promote integration of multidisciplinary data while remaining dedicated to the FAIR principles of data management (findability, accessibility, interoperability and reusability) and providing a faceted search infrastructure for locating desired data efficiently. Summaries of the datasets generated by the FaceBase projects from 2014 to 2019 are provided here. FaceBase 3 now welcomes contributions of data on craniofacial and dental development in humans, model organisms and cell lines. Collectively, the FaceBase Consortium, along with other NIH-supported data resources, provide a continuously growing, dynamic and current resource for the scientific community while improving data reproducibility and fulfilling data sharing requirements.

Fully Automatic Landmarking of Syndromic 3D Facial Surface Scans Using 2D Images.

3D facial landmarks are known to be diagnostically relevant biometrics for many genetic syndromes. The objective of this study was to extend a state-of-the-art image-based 2D facial landmarking algorithm for the challenging task of 3D landmark identification on subjects with genetic syndromes, who often have moderate to severe facial dysmorphia. The automatic 3D facial landmarking algorithm presented here uses 2D image-based facial detection and landmarking models to identify 12 landmarks on 3D facial surface scans. The landmarking algorithm was evaluated using a test set of 444 facial scans with ground truth landmarks identified by two different human observers. Three hundred and sixty nine of the subjects in the test set had a genetic syndrome that is associated with facial dysmorphology. For comparison purposes, the manual landmarks were also used to initialize a non-linear surface-based registration of a non-syndromic atlas to each subject scan. Compared to the average intra- and inter-observer landmark distances of 1.1 mm and 1.5 mm respectively, the average distance between the manual landmark positions and those produced by the automatic image-based landmarking algorithm was 2.5 mm. The average error of the registration-based approach was 3.1 mm. Comparing the distributions of Procrustes distances from the mean for each landmarking approach showed that the surface registration algorithm produces a systemic bias towards the atlas shape. In summary, the image-based automatic landmarking approach performed well on this challenging test set, outperforming a semi-automatic surface registration approach, and producing landmark errors that are comparable to state-of-the-art 3D geometry-based facial landmarking algorithms evaluated on non-syndromic subjects.

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.

Deep genotype imputation captures virtually all heritability of autoimmune vitiligo.

Autoimmune vitiligo is a complex disease involving polygenic risk from at least 50 loci previously identified by genome-wide association studies. The objectives of this study were to estimate and compare vitiligo heritability in European-derived patients using both family-based and 'deep imputation' genotype-based approaches. We estimated family-based heritability (h2FAM) by vitiligo recurrence among a total 8034 first-degree relatives (3776 siblings, 4258 parents or offspring) of 2122 unrelated vitiligo probands. We estimated genotype-based heritability (h2SNP) by deep imputation to Haplotype Reference Consortium and the 1000 Genomes Project data in unrelated 2812 vitiligo cases and 37 079 controls genotyped genome wide, achieving high-quality imputation from markers with minor allele frequency (MAF) as low as 0.0001. Heritability estimated by both approaches was exceedingly high; h2FAM = 0.75-0.83 and h2SNP = 0.78. These estimates are statistically identical, indicating there is essentially no remaining 'missing heritability' for vitiligo. Overall, ~70% of h2SNP is represented by common variants (MAF > 0.01) and 30% by rare variants. These results demonstrate that essentially all vitiligo heritable risk is captured by array-based genotyping and deep imputation. These findings suggest that vitiligo may provide a particularly tractable model for investigation of complex disease genetic architecture and predictive aspects of personalized medicine.

The genetic architecture of vitiligo.

Vitiligo is an autoimmune disease in which destruction of skin melanocytes results in patches of white skin and hair. Genome-wide linkage studies and genome-wide association studies in European ancestry cases identified over 50 vitiligo susceptibility loci, defining a model of melanocyte-directed autoimmunity. Vitiligo heritability is exceedingly high, ~2/3 coming from common and ~1/3 from rare genomic variants; ~20% of vitiligo risk is environmental. Vitiligo genetic risk is polygenic, with greater additive risk in multiplex vitiligo families than simplex cases. Vitiligo age-of-onset is bimodal, also involving a major genetic component; a MHC enhancer haplotype confers extreme risk for vitiligo (OR 8.1) and early disease onset, increasing expression of HLA-DQB1 mRNA and HLA-DQ protein and thus perhaps facilitating presentation of triggering antigens. Vitiligo triggering also involves a major environmental component; dramatic delay in vitiligo age-of-onset, especially from 1973 to 2004, suggests that exposure or response to a key vitiligo environmental trigger diminished during this period. Together, these findings provide deep understanding of vitiligo pathogenesis and genetic architecture, suggesting that vitiligo represents a tractable model for investigating complex disease genetic architecture and predictive aspects of personalized medicine.

Family Clustering of Autoimmune Vitiligo Results Principally from Polygenic Inheritance of Common Risk Alleles.

Vitiligo is an autoimmune disease that results in patches of depigmented skin and hair. Previous genome-wide association studies (GWASs) of vitiligo have identified 50 susceptibility loci. Variants at the associated loci are generally common and have individually small effects on risk. Most vitiligo cases are "simplex," where there is no family history of vitiligo, though occasional family clustering of vitiligo occurs, and some "multiplex" families report numerous close affected relatives. Here, we investigate whether simplex and multiplex vitiligo comprise different disease subtypes with different underlying genetic etiologies. We developed and compared the performance of several different vitiligo polygenic risk scores derived from GWAS data. By using the best-performing risk score, we find increased polygenic burden of risk alleles identified by GWAS in multiplex vitiligo cases relative to simplex cases. We additionally find evidence of polygenic transmission of common, low-effect-size risk alleles within multiplex-vitiligo-affected families. Our findings strongly suggest that family clustering of vitiligo involves a high burden of the same common, low-effect-size variants that are relevant in simplex cases. We furthermore find that a variant within the major histocompatibility complex (MHC) class II region contributes disproportionately more to risk in multiplex vitiligo cases than in simplex cases, supporting a special role for adaptive immune triggering in the etiology of multiplex cases. We suggest that genetic risk scores can be a useful tool in analyzing the genetic architecture of clinical disease subtypes and identifying subjects with unusual etiologies for further investigation.