Self-Exciting Population Event Models Reveal Abnormal Temporal Amplification in MAPT-Mutant Human Brain Assembloids

Human brain assembloids serve as a useful tool for investigating mutation-specific network dysfunction, yet most studies fail to differentiate between abnormal activity originating from heightened spontaneous initiation or amplified history-dependent activity. Researchers have developed a simplified discrete-time self-exciting population model to analyze calcium-imaging event sequences, which isolates three key parameters: baseline initiation (%[mu]), integrated history-dependent gain (%[eta]), and memory decay (%[tau]).

Analysis of 175 recordings from 13 MAPT p.R406W mutant or CRISPR-corrected isogenic assembloids revealed a four-fold increase in the history-dependent gain parameter (%[eta]) in the mutant samples (95% confidence interval: 3.20- 5.84; exact p=7.8x10-4), without any rise in baseline activity. Further examination showed the presence of residual Poisson overdispersion, but even with fixed-decay and beta-binomial sensitivity analyses, the mutant history effect remained intact.

The median history-dependent gain parameter (%[eta]) effectively distinguished between all 13 held-out assembloids, and a multifeature recording-level RBF-SVM achieved an area under the curve of 0.816 when evaluated on nested whole-assembloid holdout data. These findings demonstrate that robust history-dependent temporal amplification is a functional phenotype of MAPT-mutant networks.