Face scan. A large high-tech camera scans your face in 3D and – using more than 30,000 data points – extracts information from your face that you were not aware of including details of your genetic make-up. What sounds like dystopic Gattaca-like science fiction at first is actually an interesting novel technique to learn more about epilepsy-related microdeletions. It seems that some of their effects might be hidden in subtle facial features that might help understand how these genetic variants contribute to disease.
Facial recognition. Humans have a highly developed ability to recognise faces. We can tell that some people are related to each other by identifying them as “similar” without having a good idea where exactly we get this impression from. It seems like evolution has provided us with a finely-tuned facial recognition machinery that is capable to integrate a broad range of information to make sense of faces. Chinthapalli and colleagues now present a study that uses dense surface modelling of human faces to identify features that are often beyond the recognition of the human eye.
What is dysmorphic? Many genetic disorders result in particular facial features that are easily recognised and that can help clinicians make a diagnosis. The field of interpreting and cataloguing these particular features and other malformations is called dysmorphology. Some online resources refer to dysmorphology as the field of medicine that studies body characteristics that are abnormally formed. In clinical practise, we always try to avoid using the terms “normal” or “abnormal” and rather refer to “typical” and “atypical” since talking about normality always has some normative and sometimes even moral character. “Atypical”, in contrast, is purely descriptive. This difference is particularly relevant when findings are not completely disease-related but might only be associated, as in the case of microdeletion-associated facial features. This is why I like the fact that Chinthapalli and colleagues refer to atypical facial features rather than implying any sense of what is normal. The authors also mention the lack of predetermined values for a “normal face” in the paper.
What did the study show? In brief, Chinthapalli and colleagues applied a measure called Face Shape Difference (FSD) and analysed the facial gestalt of 118 patients with epilepsy that was obtained by stereophotogrammetry. 38/118 patients had pathogenic structural genomic variants (CNVs). FSD is a measure of how different a given face is when compared to a matched control cohort using the variation of facial shape. A high FSD suggests that the facial features stand out compared to the control cohort, at least statistically. On a group level, patients with CNVs had higher FSD values than controls. The authors then determined a cut-off level for optimal sensitivity and specificity and used this cut-off for a validation cohort. In the validation cohort, 4/5 patients with CNVs and 45/58 patients without CNVs were correctly identified, resulting in a sensitivity and specificity of ~80%. This translates into a positive predictive value of ~25% and a negative predictive value of > 95%. These results mean that the method is successful at discriminating to some degree between patients with and without CNVs.
FSD vs. dysmorphic features. Not all patients found to have a high FSD were considered to have dysmorphic features and there were patients with outlier FSD values with or without CNVs. This suggests that some of the features might have been subtle and not necessarily detectable by clinicians that do not have a dedicated training in dysmorphology. Also, as FSD is based on Principle Component Analysis, it is possible to find out whether the variation in facial shape in a group of patients is due to similar features. This was not the case in the study by Chinthapalli and colleagues, suggesting that the group differences were due to a collection of various distinct features rather than a common feature. Interestingly, this was also the case in patients with identical genomic variants such as the 16p13.11 microdeletion. Therefore, while the method in itself is very promising, it has not yet identified subtle “signatures” in epilepsy-related microdeletion that escape the human eye.
Facial shape and endophenotypes. Stereophotogrammetry is an interesting method for deep phenotyping or endophenotyping for epilepsy-related genetic variants. This method is objective with little possibility for human bias and might help in cases where the clinical course of a known disease is atypical. For example, the authors presented a patient with Wolf-Hirschhorn Syndrome (WHS) at the London ECE, where diagnosis was not made until adulthood. Stereophotogrammetry classified the patient as WHS-like despite the atypical clinical course. In addition, the causative gene might be difficult to pinpoint in some cases. For example, in many patients with microduplications, it is difficult to assess whether a partially duplicated gene is actually disrupted. If pathogenic variants in the gene of question are associated with a particular stereophotogrammetry signature, this method might help strengthen the case for an involvement of this gene.
Pingback: Genetic imaging in Dravet Syndrome – variation on a theme? | Beyond the Ion Channel
Pingback: Temperature rising: 17q12 microduplications and GEFS+ | Beyond the Ion Channel