🧠 neuroscience

A significant enrichment that is not: spatial nulls, co-expression, and the imaging transcriptomics of EEG alpha-power genetics

This study demonstrates that while standard spatial autocorrelation tests suggest a significant enrichment of EEG alpha-power genetics in cortical alpha generators, this finding is a methodological artifact driven by gene-set co-expression rather than biological specificity, as the signal fails to replicate across phenotypes, parcellations, and more rigorous gene-set null models.

Original authors: Schenetti, J.

Published 2026-07-25

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Schenetti, J.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine trying to understand a city by looking at a map of its traffic lights. You might notice that the lights in the downtown area all turn green at the same time, while the suburbs blink in a different rhythm. Scientists who study the brain do something similar: they look at how different parts of the brain "talk" to each other using electrical signals, like the rhythmic waves seen in an EEG (a test that measures brain activity). But here's the tricky part: they also want to know why those parts talk that way. Is it because of the specific "instructions" (genes) written in the DNA of the cells in those exact neighborhoods?

To answer this, researchers use a method called "imaging transcriptomics." Think of it as cross-referencing a map of brain activity with a library of gene blueprints. They ask: "Do the genes that make a person's brain waves strong show up more often in the specific brain regions that generate those waves?" The standard way to check this is to spin the brain map around like a globe to see if the pattern is just a lucky coincidence or if it's real. But there's a catch: genes don't work alone. They often come in families that are expressed together in the same neighborhoods, regardless of the brain activity being studied. If you only check if the map spins correctly, you might miss the fact that the genes themselves are just naturally clustered together, making it look like a special connection when it might just be a family reunion.

This is exactly the puzzle Jacopo Schenetti tackled in this study. The researcher was curious about the "alpha rhythm," a steady, humming brain wave that happens when we are awake but relaxed. It's a very famous brain signal, and we know it's highly influenced by our genetics. The big question was: Do the genes responsible for this alpha rhythm live specifically in the brain regions that generate it?

To find out, the author took a massive collection of genetic data from thousands of people and matched it with a detailed map of the brain's gene activity. They focused on the 41 specific regions known to create the alpha rhythm. Using the standard "spin test" (rotating the brain map to check for luck), the results looked promising. It seemed like the alpha genes were indeed over-represented in those specific regions, with a statistical score that suggested this wasn't just random chance. It was a neat, tidy story: the genes for the rhythm lived where the rhythm was made.

However, the author didn't stop there. They decided to run a few more tests to see if this story held up under a microscope. First, they checked if this pattern was unique to the alpha rhythm. They looked at other brain waves (theta, beta, and delta) and found something surprising: those other waves showed the same or even stronger patterns of "enrichment" in the alpha regions. If the genes were truly special to the alpha rhythm, they shouldn't be showing up so strongly for the other rhythms.

Next, they tested if the alpha genes were actually special at all. They compared the alpha gene set against 10,000 random sets of genes. The result was a "null" finding: the alpha genes were no more special than a random group of genes. The apparent "enrichment" was just a general property of how genes are clustered in that part of the brain, not a specific signature of the alpha rhythm. Finally, they changed the map they were using (the brain parcellation) to see if the result was just an artifact of how the brain was divided into regions. The "special" result disappeared when the map changed.

In the end, the paper concludes that there is no evidence that the genes for the alpha rhythm are specifically concentrated in the brain regions that generate it. The initial "positive" result was a mirage created by the way genes naturally cluster together, a trap that the standard spin test failed to catch. The author suggests that future studies need to use a different kind of "null test" (checking against random gene sets) to avoid these false alarms. While the alpha rhythm is definitely heritable, this study shows that we can't yet point to a specific molecular neighborhood in the brain that is uniquely responsible for it. The search for the genetic home of the alpha rhythm continues, but this particular map turned out to be a bit of a trick.

Technical Summary: "A significant enrichment that is not: spatial nulls, co-expression, and the imaging transcriptomics of EEG alpha-power genetics"

Problem Statement
The study addresses a critical gap in imaging transcriptomics: determining whether the genetic architecture of a specific neurophysiological trait is spatially organized in the brain regions that generate that trait. Specifically, the authors investigate whether genes associated with EEG alpha power are over-expressed in the 41 cortical regions identified as alpha-rhythm generators (Tabarelli et al., 2022).

Methodologically, the paper targets a known limitation in the field's standard practice. Enrichment analyses typically rely on a "spin test" (spatial permutation) to control for spatial autocorrelation. However, this approach does not account for the fact that genes within a set are often co-expressed, creating spatial structure independent of the trait of interest. The authors hypothesize that a result appearing significant under a spatial null might be a generic property of arbitrary gene sets in that cortical territory rather than a trait-specific signal.

Methodology
The study utilizes a multi-stage pipeline integrating GWAS summary statistics, regional gene expression, and rigorous null modeling:

Data Sources:
- Genetics: ENIGMA-EEG summary statistics (N ≈ 6,000–7,700) for six phenotypes: central alpha power (primary), occipital alpha power, occipital alpha peak frequency, and theta, beta, and delta power.
- Transcriptomics: Regional gene expression from the Allen Human Brain Atlas (AHBA), processed via abagen and mapped to the Glasser HCP-MMP1.0 atlas (360 regions), with sensitivity analyses in Schaefer-100 and Yan-600 parcellations.
- Target Regions: 41 cortical alpha-source regions (bilateral, 82 parcels) defined by Tabarelli et al. (2022) using MEG source reconstruction.
Gene-Based Analysis:
- MAGMA was used to compute gene-level association statistics.
- Two enrichment scores were defined per region: a continuous score (Spearman correlation between regional expression and genome-wide gene z-statistics) and a top-100 score (mean expression of the top 100 genes ranked by p-value).
Null Models and Validation:
- Spatial Null (Spin Test): 10,000 spatial rotations of parcel centroids on the cortical sphere to preserve spatial autocorrelation.
- Gene-Set Null: 10,000 random gene sets of matched size to test if the observed enrichment is specific to the alpha-associated genes or a property of any co-expressed set.
- Positive Control: Testing the pipeline against the first principal component (PC1) of the AHBA expression matrix (a known cortical gradient) to ensure sensitivity.
- Surrogate Validation: Cross-checking the custom spin implementation against brainsmash surrogates.

Key Results

Initial Spatial Significance: Under the field-standard spin test alone, the study initially appeared to support the hypothesis. Central alpha power showed significant enrichment in the alpha-source regions (continuous $p_{spin} = 0.022$ ; top-100 $p_{spin} = 0.030$ ). This result was corroborated by the brainsmash surrogate ( $p=0.018$ ) and the positive control ( $p_{spin} = 2 \times 10^{-4}$ ).
Failure of Specificity: The enrichment was not band-specific. Theta, beta, and delta power showed comparable or stronger enrichment in the alpha-source regions. Furthermore, the result did not replicate across alpha phenotypes; occipital alpha power and alpha peak frequency were non-significant.
Gene-Set Null Dissolution: When assessed against random gene sets of matched size, the alpha top-100 set was indistinguishable from random ( $p_{geneset} = 0.33$ ). The observed enrichment was a generic property of arbitrary gene sets in the occipito-parieto-frontal territory, invisible to the spatial-only null.
Lack of Robustness: No test survived False Discovery Rate (FDR) correction across the 24-cell analysis grid (minimum $q = 0.127$ ). Significance was not robust to changes in parcellation; it was only nominal in the native Glasser space and disappeared or weakened in Schaefer-100 and Yan-600.

Key Contributions

Substantive Finding: The study provides evidence that EEG alpha-power genetics does not show a regionally specific transcriptomic signature in the cortical generators of the alpha rhythm. Any weak tendency toward enrichment is shared across frequency bands and is likely driven by the general genetic correlation of EEG power rather than an alpha-specific molecular architecture.
Methodological Demonstration: This work serves as a "worked demonstration" that correcting for spatial autocorrelation is necessary but insufficient. It shows that a result can be spin-significant, surrogate-corroborated, and mechanistically plausible, yet be entirely accounted for by gene-set co-expression.
Proposed Standard: The authors argue that enrichment claims in imaging transcriptomics must report a gene-set null alongside the spatial null to avoid false positives driven by co-expression confounds.

Significance and Claims
The paper claims that the apparent enrichment of alpha-associated genes in alpha generators is a statistical artifact of co-expression structure rather than a biological truth. The authors emphasize that the "null result" is robust because the study design was built to resist over-interpretation, utilizing internal power-matched controls, positive controls, and multiple null models.

The methodological conclusion is presented as the primary generalizable takeaway: relying solely on spatial nulls (like the spin test) can lead to the misattribution of generic cortical expression patterns to specific trait genetics. The authors conclude that future enrichment claims should treat results passing only the spatial null as uninterpreted, and that the cost of adding a gene-set null is negligible compared to the risk of a literature populated by properties of the cortex rather than the genes.

The study does not claim to have found a new biological mechanism for alpha power but rather to have rigorously tested and refuted a specific hypothesis regarding its regional transcriptomic organization, while simultaneously exposing a methodological vulnerability in the field.

Technical Summary: "A significant enrichment that is not: spatial nulls, co-expression, and the imaging transcriptomics of EEG alpha-power genetics"

More like this