Early detection of diseases is a central challenge for pediatric medicine. The earlier a disease is discovered, the easier it is to avoid complications and sequelae and to reduce long-term morbidity. This is particularly relevant for children with rare diseases in whom the diagnostic process is often delayed. Children with rare and chronic diseases are usually only diagnosed when their disease manifests or complications arise. Thus, there is an urgent need to develop and use new sensitive and specific diagnostic methods, preferably as non-invasive as possible.
Next generation sequencing technologies have revolutionized human genetics. A growing number of hereditary monogenic rare diseases has been identified, leading to a better understanding of molecular processes even in multifactorial diseases. In addition to genomics, other omics-technologies (e.g. transcriptomics, metabolomics, proteomics, immunomics) complement our scientific armamentarium to comprehensively assess states of diseases. A challenge of these technologies is to integrate and interpret these large datasets. Emerging data suggest that combining multi-layer omics data with digital clinical data will allow us to improve diagnostics, to optimize prevention, and to design definitive cures. Advances in machine learning enabling pattern recognition and statistical associations, offer new perspectives for developing innovative and non-invasive diagnostic methods.
In the context of this non-randomized, monocentric observation study, the benefit of using a combination of pattern recognition of image data of the retina by fundus photography and optical coherence tomography (OCT) in combination with the analysis of various OMICS data (genome, transcriptome, proteome and metabolome) will be explored in search of markers for rare and chronic childhood diseases. Retinal images and OMICS data are pseudonymized and subjected to machine learning algorithms. Starting from classical nosological entities, we will compare the data not only within defined groups but also across phenotypes, aiming to shed light on pleiotropic factors. Once associations between genomic and phenotypic data sets become apparent, new hypotheses will be developed and tested in suitable model systems.
