Can't diagnose neurodevelopmental disorders? AI to help

Recently, Mohamed Uddin and Marc Woodbury Smith from the center of applied genomics of children's Hospital in Toronto, Canada, and Yujiang Wang from the Institute of neuroscience, University of Newcastle, published an article on the application of artificial intelligence (AI) in precision medicine for neurodevelopmental disorders in the sub column npj digital medicine of nature 。 Precision medicine is a medical approach, which uses a variety of technologies to guide the diagnosis and treatment tailored for patients The availability of technologies, including high-performance computing (HPC) and large-scale biological data sets, is critical to achieving a precise healthcare approach that impacts healthcare The core of the strategy is a set of computer algorithms, which are used to identify the patterns in the multi-dimensional data set, and then predict or optimize according to the availability of similar data of a single patient Artificial intelligence algorithm applies the learning strategy based on classification or pattern recognition to (multidimensional) input data, so as to be able to predict from the future data set For example, in clinical medicine, it may involve using the results of pathological specimens to predict the diagnosis and treatment of pathological specimens of new patients Neurodevelopmental disorder (NDD) begins in early childhood and affects many functional areas, including cognitive and executive functions, language and social functions, motor functions and behavioral control Many different diagnoses fall into this category, including mental disability (ID), autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD), tic disorder, and other motor disorders Epilepsy and other early-onset brain diseases are also classified as NDD Autism and ADHD are common lifelong diseases, and men are more vulnerable than women In contrast, some syndromic NDD, especially monogenic disease, are very rare In patients with different degrees of severity, two or more NDD diagnosis of comorbidity is common In addition to some epileptic syndromes (such as Western syndrome) can be diagnosed more clearly according to the results of EEG, most NDD, including epilepsy itself, are diagnosed according to the symptoms determined by direct observation or family history This is particularly problematic because the validity of reliable information varies from individual to individual, and even expert opinion varies from clinician to clinician, so diagnostic recognition is often not decisive In addition, due to the nature of the development of such diseases, the clinical situation may be different over time, and the symptoms will become more or less serious as the children grow up It is obvious that a more stable and objective method is needed to classify individuals with NDDs, but at present, the fuzziness of this diagnostic approach is an important obstacle for accurate medical treatment of neurodevelopmental disorders Functional magnetic resonance imaging (fMRI), which can capture brain activity with high resolution, is an important index in the diagnosis and treatment of NDD in the 1990s Since then, the human genome has been mapped, exome and whole genome sequencing techniques have led to the detection of hundreds of pathogenic genes and loci for ASD and other NDD In fact, exome or genome sequencing is becoming more and more frequent and cost-effective for newborns with high risk of genetic abnormality Subsequently, the emergence of transcriptome sequencing dependent technology established the Allen development human brain map in 2011, analyzed the coding database of non coding elements in the human genome in 2012, and coded the human cell map in 2017 During 2012 and 2014, a number of sequencing consortia focusing on NDD were also launched to identify disease-related variants and make the exon set and WGS data available for further research by the scientific community Considering that most of the identified genetic variations have unknown pathogenicity and little is known about the functional consequences, the discovery of CRISPR / CAS as a gene editing tool in 2012 enables scientists to better describe the identified genetic variations In recent years, AI has been used in autism, epileptic encephalopathy, mental disorders, attention deficit hyperactivity disorder and rare genetic diseases When AI in NDD is discussed in this article, three layers of analysis will be considered The most important topic is to apply these methods to multidimensional NDD biological datasets, as well as their complexity AI is very important to identify causal genes and loci Although the effective rate of genetic diagnosis of severe syndrome is about 50%, the gene or locus of NDD is still unknown AI has recently made great achievements in improving the genetic diagnosis of NDD For example, deepsea is a kind of depth algorithm, which can predict the effect of non coding variables from the beginning of sequence data The model uses large-scale chromatin analysis data, including transcription factor binding, DNase-I sensitivity, and histone labeling to predict the functional consequences of noncoding mutations The application of AI in NDD has the complexity of deep type and large-scale histological data If AI will completely change the medical services of NDD patients, there are still some major obstacles to be solved For example, the regulatory role of the environment, in terms of technology, due to the different methods, tools and protocols used to collect data, and so on AI is at the forefront of therapeutic interventions and drug design Currently, there are 51 kinds of targeted gene specific drugs approved by the food and Drug Administration (FDA) for neurological and psychiatric diseases The emergence of sequencing technology mainly focuses on promoting the implementation of early accurate diagnosis Accurate treatment is still a challenge for NDDs Recently, the emergence of genomic editing technology (CRISPR / cas9) and antisense oligonucleotide therapy has enabled scientists to simulate cell phenotypes and help identify precise molecular targets For example, the application of CRISPR / cas9 can help to knock out the functional copy of chd8 gene in induced pluripotent stem cells (iPSCs) The knockout iPSCs showed the differential expression of thousands of genes in neural progenitor cells and affected the early differentiated neurons 99crispr / cas9 or other CAS family proteins are still prone to errors, and the success rate of the experiment is not high The future hope is that CRISPR / cas9, antisense oligonucleotide therapy and gene therapy based technology will enable us to detect the precise target molecules of most mutated genes in NDD This will eventually lead to the design of target molecules (i.e antisense oligonucleotides or siRNA) to inhibit or destroy the experimental pathway of the wrong pathway This kind of drug design will need to promote the realization of artificial intelligence algorithm AI is already affecting health care, and researchers hope that some of the successes achieved so far in cancer and cardiovascular disease will also be seen in NDD Significant investments are needed in new treatments that map to different disease categories and subclasses, and researchers need to stay away from existing diagnostic structures to reach more intermediate biologically driven phenotype levels that map more clearly to treatment responses and clinical outcomes The healthcare sector is faced with an arduous task: to develop infrastructure, to translate professional knowledge among healthcare professionals, and to participate in the development of new clinical pathways with service users themselves The short-term investment will definitely have long-term benefits, which can not only save the financial expenses brought by precision medical treatment, but also ultimately improve the health status of the people Reference source: https://www.nature.com/articles/s41746-019-0191-0 Jenny turtle statement: this point of view only represents the author, not the position of yaozhi.com, welcome to exchange and supplement in the message area; if you need to reprint, please be sure to indicate the author and source of the article.