Nature's Whole-Brain Association Reproducibility Study

By | November MRI has changed our understanding of the brain by mapping some cognitive-behavioral abilities to the structure and function of specific brain regions
.

But mental health research has not been able to make similar progress from MRI data, largely because it is difficult to link brain structure and function to complex cognitive abilities or mental health phenotypes across individuals
.

Brain-wide association studies (BWAS) aim to predict disease characteristics by quantifying the association between behavior and phenotype
.

However, the current association studies have too few sample data and insufficient reproducibility
.

Recently, Nico UF Dosenbach's research group at Washington University School of Medicine, Damien A.
Fair's research group at University of Minnesota School of Medicine, Scott Marek at Washington University School of Medicine (first author), and Harvard Medical School Brenden Tervo-Clemmens (first author) The collaborative paper, titled Reproducible brain-wide association studies require thousands of individuals, established large-scale reproducible brain-wide association studies and found that at least thousands of samples are required to ensure reproducible brain-wide association studies
.

Whole-brain association studies are necessary to predict and reduce the burden of psychiatric disease and improved human intelligence and cognitive function
.

However, the acquisition of MRI data is very expensive, resulting in the inability to replicate the results of whole-brain association studies with small samples, and the shortcomings of population-based psychology and genomics studies such as insufficient statistical power
.

Small studies are most susceptible to sampling variability, the random variation in associations between population subsamples
.

As the sample size increases, the sampling variability decreases and the association gradually stabilizes
.

Therefore, if we want to truly achieve whole-brain associations, larger sample sizes are needed for confirmation
.

To this end, the authors employed three of the largest neuroimaging datasets currently available, with a total sample size of approximately 50,000 individuals, to quantify BWAS effect size and reproducibility as a function of sample size
.

The three databases are the Adolescent Brain Cognitive Development (ABCD) database, which includes 11,874 samples; the Human Connectome Project (HCP), which includes 1,200 samples, and the UK Biobank database (UK Biobank).
, UKB) including 35,735 samples [1-3]
.

In order to be able to accurately estimate the size of the brain-wide association study effect, the authors' strategy was to start the study with the Adolescent Brain Cognitive Development Research database and to validate the resulting association effect using the Human Connectome Project and the UK Biobank database
.

The authors correlated brain characteristics such as cortical thickness and resting-state functional connectivity (RSFC) with 41 analysis measures for demographic, cognitive ability, and mental health measures
.

In order to reduce the influence of variables such as head movement on the association analysis, the authors used strict quality control to analyze the cerebral cortex and two widely studied phenotypes (cortical thickness and RSFC), cognitive ability scores and associations with Psychopathological Checklist scores
.

The effect with the largest association had a median |r| value greater than 0.
06, and the top 10% of the largest associations were distributed in sensorimotor and cortex
.

Figure 1 Whole-brain association studies: associations between structure and function require large-scale data to facilitate reproducibility of associations The authors found that the error rate in the statistics depends on the magnitude of the effect of the association tested and the threshold for significance testing, Even with more than 1,000 test samples, the false-negative rate is very high, and half of the statistically significant associations are exaggerated
.

Therefore, in order to obtain a true association study, it is necessary to expand the sample size and reduce the false negative rate
.

In addition, the authors found stronger correlations using multivariate brain-wide association studies than univariate methods
.

In behavioral phenotypes, multivariate sample associations are more robust than univariate effects
.

There is no one-size-fits-all research protocol, and small sample neuroimaging studies are still important
.

By repeatedly sampling the same individual, precise individual-specific data can be generated
.

In addition, whole-brain association studies also benefit from the establishment of multiple databases.
Data-driven, multi-dimensional combined brain behavioral phenotypes will further advance our understanding of cognitive ability and mental health
.

In general, the work found that brain-wide association studies between neuroimaging databases such as MRI and human behavior require larger sample sizes, at least thousands of data samples to ensure the authenticity of the association
.

But fine-grained neuroimaging studies in small samples remain critical for behaviors such as mental health
.

Multi-database and multi-dimensional research will better provide a clearer prospect for the correlation research of complex human behavior
.

Original link: https://doi.
org/10.
1038/s41586-022-04492-9 Publisher: Eleven References 1.
Casey, BJ et al.
The Adolescent Brain Cognitive Development (ABCD) study: imaging acquisition across 21 sites 32, 43–54 (2018).
2.
Van Essen, DC et al.
The WU-Minn Human Connectome Project: an overview.
NeuroImage 80, 62–79 (2013).
3.
Sudlow, C.
et al.
UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
PLoS Med.
12, e1001779 (2015).
Reprint instructions [Original article] BioArt original article, welcome Personal forwarding and sharing is prohibited without permission.
The copyright of all published works is owned by BioArt
.

BioArt reserves all legal rights and violators will be held accountable
.