confirmed our main findings with multiple external scRNA datasets using FUMA. This provides further evidence that genetic underpinnings of neurological disorders are distinct from those of psychiatric, SUDs, and behavioral/quantitative phenotypes [8, 10]. Our main findings were based on the identification of cell types by LDSC, DEPICT, and MAGMA top 10% mode. MAGMA linear mode was omitted because its strength of association estimates were consistently deviating substantially from LDSC, DEPICT, and MAGMA top 10% mode. Therefore, it was deemed too lenient and thus prone to type I error inflation. This concurs with previous studies reporting that binned MAGMA analyses in linear mode inflated results since the binned scores can have strong correlations with the average gene expression across cell types [29]. Also in agreement with previous lines of evidence, we confirm that the statistical foundation of the SNP-wise mean gene analysis model MAGMA < 1.07 may result in biased associations of cell types [11].
