Assessment of Autism Spectrum Disorder Severity - Dr. Roser

Abstract: Autism Spectrum Disorder (ASD) is an incredibly heterogeneous neurodevelopmental disorder with strong genetic component. ASD severity is currently defined by the amount of support needed by the subjects, a measure not easily quantifiable. Understanding and quantifying ASD severity are key for diagnosis, treatment planning, and prognostic predictions. The contribution of non-coding variation to ASD severity remains mostly unexplored. The goal of this study was to detect non-coding variants related to ASD severity in individuals of the Simons Simplex Collection. Using unsupervised methods with 21 behavioral tests in 2857 individuals, we were able to define groups of low and high ASD severity. We then used whole-genome-sequencing data for 278 of those individuals to evaluate the contribution of non-coding variation to severity. We focused on a subset learning-regulated non-coding regions, significantly associated with known ASD risk genes. The binary classification for severity obtained from the phenotypic analysis was used for selection of SNPs and CNVs via Elastic-Net. The parameters were tuned by 10-fold cross-validation. The final set of parameters showed an area under the ROC curve of 0.733 ± 0.02. Using these parameters a total of 191 variants were selected. The variants mapped to 198 genes. The gene list includes several important components of the synapse, as mGluR5, whose dysregulation has been associated with ASD. Overall, our preliminary studies show that ASD severity can be quantified using a combination of phenotypic traits, and that genetic variants located in learning-regulated regions are useful for prediction.

Speaker: Leandro Roser, Ph.D., is a postdoctoral associate at Elson S. Floyd College of Medicine, Washington State University. He received his PhD from the University of Buenos Aires, Argentina. During his PhD, he combined population genetics and ecology, and developed an R package for Landscape Genetics (https://cran.r-project.org/package=EcoGenetics). He also has other packages in scientific repositories (https://cran.r-project.org/package=chunkR, https://bioconductor.org/packages/FastqCleaner). He is currently working in Autism Spectrum Disorder research from a Machine Learning viewpoint, developing pipelines and software for analysis of genomics information, useful for precision medicine research. He is also involved in other research projects, analyzing transcriptomics data.