Computational axis
Multiscale forward and inverse modelling will connect local spiking circuit dynamics, neural mass models, EEG/MEG features, and simulation-based inference.
About
Project Overview and Scientific Rationale
Autism spectrum disorder shows substantial heterogeneity across genetic, neural, cognitive, behavioural, clinical, and developmental levels. MechNeuro-ASD addresses this heterogeneity by linking cortical circuit mechanisms to EEG-derived biomarkers and richly characterized longitudinal paediatric data.
Current diagnostic categories often mask biologically meaningful variability. By focusing on neural circuits, MechNeuro-ASD aims to identify interpretable axes along which autistic children can be meaningfully stratified.
Clinical axis
A multisite paediatric cohort will combine clinical, neuropsychological, contextual, resting-state EEG, and task-evoked EEG measures across follow-up.
Neuroethical axis
Procedures will be shaped by neurodiversity-affirming principles, family-centred priorities, and lived-experience perspectives on paediatric EEG and biomarkers.
- Circuit-level biomarkers: EEG and MEG provide millisecond-scale access to neural dynamics, but their mechanistic specificity must be validated.
- Simulation-based inference: Biophysical simulations and Bayesian inverse methods can estimate hidden circuit parameters from observed signals.
- Developmental context: Longitudinal, context-rich paediatric data are essential for robust and clinically meaningful biomarkers.
- Responsible translation: Biomarker development must avoid stigma and remain aligned with neurodiversity-affirming care.
General Objectives
GO1 Computational
Advance tools for inferring mechanistic relationships between cortical circuit parameters and electrophysiological biomarkers using LFP, EEG, and MEG signals.
GO2 Clinical
Establish a large, ecologically valid, and phenotypically diverse cohort of pre-adolescent autistic children with longitudinal multimodal follow-up.
GO3 Translational
Bridge model-derived neural mechanisms with clinically meaningful outcomes to support neural stratification, prognosis, and intervention monitoring.
GO4 Bioethical
Ensure procedures involving autistic children are respectful, inclusive, neurodiversity-affirming, and aligned with family-centred priorities.
Work Packages
The work plan combines computational development, mouse LFP validation, multisite paediatric recruitment, multidimensional clinical characterization, EEG acquisition, translational modelling, and bioethical training.
- WP1-WP4: Multiscale modelling, feature selection, robust SBI, and pre-clinical validation using mouse LFP data.
- WP5-WP7: Recruitment, multidimensional assessment, and resting-state and task-evoked EEG in autistic children.
- WP8: Linking EEG-derived mechanistic parameters to clinical, behavioural, and contextual outcomes.
- WP9-WP10: Bioethics, neurodiversity-affirming training, scientific communication, and public outreach.
Methods
Forward modelling
Local spiking circuit models, kernel-based extracellular signal prediction, realistic head models, and neural mass simulations will generate biologically plausible EEG/MEG datasets.
Inference and feature selection
Simulation-based inference, feature-selection methods, multi-fidelity strategies, and robustness-to-misspecification approaches will be evaluated for realistic circuit models.
Clinical data integration
Linear mixed-effects models and normative modelling will integrate mechanistic EEG estimates with clinical, behavioural, neuropsychological, contextual, and developmental variables.
Impact and Responsible Practice
Scientific and technical impact
The project will advance systems neuroscience and autism biomarker science through validated modelling tools, inference pipelines, paediatric EEG workflows, and open computational resources.
Social and clinical impact
Neurodiversity-affirming procedures, family engagement, and mechanistically interpretable biomarkers will support earlier identification of support needs and more individualized care pathways.
Data and inclusion
Computational outputs will follow open-science principles, while clinical data will be protected under strict privacy safeguards. Sex and gender will be modelled explicitly to address underrepresentation of autistic girls.