# A mega-analysis framework for delineating autism neurosubtypes

> **NIH NIH R01** · CHILD MIND INSTITUTE, INC. · 2024 · $754,050

## Abstract

ABSTRACT
This application proposes to lay the groundwork for precision medicine approaches to autism spectrum disorder
(ASD) by identifying reproducible clinically relevant brain-connectome-based subtypes. The proposal addresses
the clinical and biological heterogeneity of ASD by focusing on the intermediate level of analysis of systems
neuroscience, following clues that ASD is associated with abnormalities in the brain functional connectome.
Thus, we aim to identify neurosubtypes (NS), i.e., subgroups of individuals with homogeneous atypical features,
based on measures of intrinsic functional connectivity (iFC). Primary aims are to: 1) generate a large,
retrospectively harmonized data resource with comprehensive assessment of iFC and clinical phenotypes; 2)
identify iFC-based neurosubtypes and establish their associations with clinically relevant phenotypes; 3) test the
replicability of neurosubtypes and their associations with phenotypic measures in an independent sample . To
this end, we propose to leverage existing large-scale ASD neuroimaging data collections from the Autism Brain
Imaging Data Exchange, the National Database for Autism Research, and the Healthy Brain Network. Sample:
Age/Sex: Boys and girls, 6-18 years old. Diagnosis: ASD and neurotypical (NT) individuals. Size: to date, the
above neuroimaging resources contain a total N=3528; ASD n=2136, NT n=1392. Methods: Following
systematic and extensive data organization, rigorous quality assurance, and preprocessing we will proceed with
quantitative data harmonization using state-of-the-art methods. CovBat, the most advanced version of the
Bayesian framework, ComBat, will be applied to harmonize MRI data. It has been developed by Co-I Shinohara
to control for inter-scanner differences in MRI-based measures, as well as for errors arising from subject
differences in measurement covariance. Recent advances in item response theory will be used to harmonize
phenotypic data, informed by preliminary clinical work. To further enhance our clinical data harmonization efforts,
the neuroimaging data will be aggregated with phenotypic-only collections from Co-Is Lord and Bishop (ASD
n=1513). Connectopathy features: To scope the entire spectrum of ASD connectopathy, multiple features will be
assessed simultaneously for the first time. Neurosubtypes: Building on our feasibility work with Co-I Yeo,
homogeneous neural ASD subgroups will be identified through novel Bayesian latent factor modeling. It allows
for subjects to belong to subtypes in varying degrees, identifying hybrid, categorical and dimensional,
neurosubtypes. Other key questions include the relevance of MRI features studied, the diagnostic specificity of
neurosubtypes, and cross-subtyping method validity. The neurosubtypes identified and methods for
harmonization, along with all data generated for mega-analyses will be regularly shared, starting at the end of
year two. Findings will address critical knowledge gaps and the novel resource ...

## Key facts

- **NIH application ID:** 10817940
- **Project number:** 5R01MH133334-02
- **Recipient organization:** CHILD MIND INSTITUTE, INC.
- **Principal Investigator:** Adriana Di Martino
- **Activity code:** R01 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $754,050
- **Award type:** 5
- **Project period:** 2023-04-01 → 2028-01-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10817940

## Citation

> US National Institutes of Health, RePORTER application 10817940, A mega-analysis framework for delineating autism neurosubtypes (5R01MH133334-02). Retrieved via AI Analytics 2026-05-27 from https://api.ai-analytics.org/grant/nih/10817940. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*