Delta power robustly predicts cognitive function in Angelman syndrome

Abstract

Objective: Angelman syndrome (AS) is a severe neurodevelopmental disorder caused by loss of function of the maternally inherited UBE3A gene in neurons. Promising disease-modifying treatments to reinstate UBE3A expression are under development and an early measure of treatment response is critical to their deployment in clinical trials. Increased delta power in EEG recordings, reflecting abnormal neuronal synchrony, occurs in AS across species and correlates with genotype. Whether delta power provides a reliable biomarker for clinical symptoms remains unknown.

Methods: We analyzed combined EEG recordings and developmental assessments in a large cohort of individuals with AS (N = 82 subjects, 133 combined EEG and cognitive assessments, 1.08-28.16 years; 32F) and evaluated delta power as a biomarker for cognitive function, as measured by the Bayley Cognitive Score. We examined the robustness of this biomarker to varying states of consciousness, recording techniques and analysis procedures.

Results: Delta power predicted the Bayley Scale cognitive score (P < 10^-5 , R² = 0.9374) after controlling for age (P < 10^-24 ), genotype:age (P < 10^-11 ), and repeat assessments (P < 10^-8 ), with the excellent fit on cross validation (R² = 0.95). There were no differences in model performance across states of consciousness or bipolar versus average montages (ΔAIC < 2). Models using raw data excluding frontal channels outperformed other models (ΔAIC > 4) and predicted performance in expressive (P = 0.0209) and receptive communication (P < 10^-3 ) and fine motor skills (P < 10^-4 ).

Interpretation: Delta power is a simple, direct measure of neuronal activity that reliably correlates with cognitive function in AS. This electrophysiological biomarker offers an objective, clinically relevant endpoint for treatment response in emerging clinical trials.

FIGURE 1

EEG data processing procedure. (A) Data were referenced to three standard montages for review and analysis: the linked‐ear, bipolar, and average references. Electrodes used in computing the reference are highlighted in blue. (B) EEG data were manually reviewed and staged for periods of sleep and wakefulness. (C) Data were visually inspected and manually cleaned to remove non‐cerebral artifacts (highlighted here in yellow). An example “clean” dataset is shown (bottom row) with breakpoints identified by vertical black lines. (D) Power spectral densities (PSDs) were calculated for each EEG recording. The delta [alpha] frequency range is indicated with a blue [red] rectangle on the plotted PSD. An example EEG trace (grey) showing filtered delta [alpha] activity in blue [red] in a sample EEG trace is shown in the inset.

FIGURE 2

Delta power predicts cognitive function. (A) Both log₁₀(age) (P < 10⁻³¹) and genotype:log₁₀(age) interaction (P < 10⁻¹²) predict cognitive function. Each circle indicates a subject visit. The solid [dashed] curves indicate model fit [95% confidence intervals]. Subjects with deletion [non‐deletion] genotype are indicated in red [blue]. (B) Amongst longitudinal subjects (N = 27 subjects, N = 51 pairs), change in delta power predicts change in the Bayley Cognitive Score (R ² = 0.0814, P = 0.0386, β = −13.40 [SE 6.30]). (C) Subjects had similar cognitive scores across longitudinal assessments, with different baseline values (e.g., y‐intercepts). Connected points in color indicate longitudinal data from the same subject. Points in gray indicate subjects with only one visit. (D) The observed Bayley Cognitive Scores and the Bayley Cognitive Scores predicted by the model were highly correlated (R ² = 0.9374), indicating a good model fit. Each circle indicates a real and predicted subject score. The dashed line indicates identical values between the two scores. (E) There is a linear relationship between delta power and the Bayley Cognitive Score after controlling for age, age: genotype interaction, and repeat visits using the mixed‐effects model shown at the top of the panel (R ² = 0.9374, delta power P < 10⁻⁵, log10(age) P < 10⁻²⁴, genotype:log₁₀(age) P < 10⁻¹¹, repeat subjects P < 10⁻⁸). The solid black line shows the model fit and the gray shaded region indicates the 95% confidence intervals. Red circles indicate combined EEG and cognitive score visits. Red lines indicate repeat visits by the same subject. The insert shows the standard 10–20 EEG channels used to estimate delta power. *The Bayley Cognitive Score has been adjusted for a fixed age and genotype for visualization.

FIGURE 3

Model validation. (A) The residuals are normally distributed, with a low root mean squared prediction error (RMSPE = 2.50). (B) there is a strong correlation (R ² = 0.9455) between the actual and predicted Bayley Cognitive Score values for each excluded subject, indicating a good model fit. Each circle indicates a subject visit and colors represent the same subject in (A) and (B).

FIGURE 4

Sub‐sampled electrode clusters and raw EEG data predict cognitive function in AS. Sub‐sampled clusters of electrodes are shown in the first row (blue = channel included in montage), with the groupwise abbreviation and the names of channels included listed in the second row. The full average reference is shown in the second column in gray for comparison, and cleaned data in the full average reference, to which all other cells are compared, is shown in dark gray. For each of the sub‐sampled channel montages, and the full average reference, relative delta power was computed using cleaned, raw, and raw data with balanced duration to cleaned data, and used to predict the Bayley Cognitive Score. Cells with AIC scores superior (ΔAIC > 4) to those from the model using cleaned data from all electrodes in the full average reference are highlighted in green.

FIGURE 5

Delta power calculated from raw data in the OP montage predicts cognitive function. There is a linear relationship between delta power, estimated from raw data in the OP montage (inset), and the Bayley Cognitive Score after controlling for age, age:genotype interaction, and random effects for repeat subjects (R ² = 0.9517, delta power P < 10⁻⁸, log₁₀(age) P < 10⁻²², genotype:log₁₀(age) P < 10⁻¹², repeat subjects P < 10⁻⁸). The solid black line indicates the linear model fit, and the gray shaded region indicates the 95% confidence interval. Longitudinal same‐subject data are connected by red dotted lines. *The Bayley Cognitive Score has been adjusted for the impacts of age and genotype and the recast values plotted.

FIGURE 6

Motor and language function are predicted by delta power with reduced data collection criteria. Delta power estimated using the OP regional reference montage (insets), and raw EEG data was used to predict the (A) Bayley Receptive Communication Score, (B) Bayley Expressive Communication Score, (C) Bayley Fine Motor Score, and (D) Bayley Gross Motor Score. For all plots, the solid black line shows the linear fit, and the gray shaded region indicates the 95% confidence interval. Longitudinal same‐subject data are connected by red dotted lines. The R ² value given by the model and the P value for delta power are shown in bolded text. *Scores have been adjusted for a fixed age and genotype for visualization.