Discovering Molecular and Neural Biomarkers of Social and Language Development in ASD Toddlers

点击翻译按钮获取中文摘要

Social affect, social processing, and social communication symptoms differing in degrees of severity appear in the first years of life in autism spectrum disorder (ASD), giving rise to pervasive, lifelong challenges to patients, parents, and therapists. Lack of verified early-age biological and clinical ASD subtypes prevents accurate prediction of clinical progression and treatment outcome. ASD social affective symptoms are ASD specific signs and can be indexed by motherese eye tracking (ET); social differences are undoubtedly caused by neural dysfunction such as those found by our grant, but the molecular pathobiology underlying neural and ET social heterogeneity are not well understood. Subtypes of fMRI neural responses to social affective language are known from our grant work, as are atypical cortical patterning for a subgroup of ASD toddlers with poor language and social outcomes. Our recent work identifies gene expression features associated with such subgroups, and dysregulated signaling pathways are related to social symptom severity. Missing is how these various imaging, social and gene findings are connected in ASD subtypes, and what molecular drivers underlie social neural activity. AIM 1 - discover underlying molecular drivers of ASD social neural and social ET subtypes. Using an unsupervised data-driven precision medicine method (Similarity Network Fusion (SNF)) for patient subtyping, we will integrate multimodality social clinical, social eye tracking, social speech activation fMRI, MRI-based cortical parcellation, and transcriptomic data from a large new cohort of toddlers to discover molecular dysregulations underlying subtypes of ASD social neural dysfunction. Replication analyses will use our existent NIDCD cohort as an independent sample. Subtypes will be tested for robustness and reproducibility. Features in each modality that differentiate ASD subtypes will be identified via tests of each pair of subtypes over each data modality. AIM 2 - prenatal molecular processes and cell types associated with ASD neural-social subtypes. Using normative BrainSpan expression data to map prenatal temporal-spatial activity of pathways disrupted in ASD subtypes, we will learn when and where disrupted pathways in ASD subtypes are normally upregulated during prenatal development. Top differentially expressing (DE) genes in each disrupted pathway will be analyzed for enrichment in GO biological processes. Using normative cell type- specific gene markers, we will test whether DE genes in ASD neural-social subtypes enrich prenatal and postnatal cell type-specific markers, and determine which neural-social subtypes strongly express a recently identified multi-pathway DE-ASD network that is over-active in ASD neural progenitors and neurons. AIM 3 – prognosis prediction from ASD subtypes. For toddlers in each ASD subtype, we use Bayesian multimodality sparse linear regression learning to determine which transcriptomic, brain imaging, eye tracking and treatment features in each subtype predict each child's outcome social and language 1.5 years after clinical intake. This opens avenues for clinical actionable prognostic outcome subtype markers for individual patients.