(A) Two-dimensional grid searches over free model parameters. For each heatmap, gain modulation is scaled in proportion to regional expression levels of different serotonergic (HTR) and dopaminergic (DRD) receptor-encoding genes, each of which is agonized by LSD. Black stars indicate maximal model-empirical loadings for each heatmap. (B) Model-empirical loading (left axis) and Spearman rank correlation (right axis) are greatest when gain is modulated by regional expression levels of HTR2A. Model change in GBC (ΔGBC) maps used in this analysis were generated using the gain-modulatory parameters that maximized model-empirical loadings, as indicated on each heatmap. (C,D) Following Burt et al., 2020, we generate surrogate brain maps with randomized spatial topographies that, by construction, exhibit spatial autocorrelation that has been matched to that of the HTR2A map. These spatial autocorrelation-preserving surrogate brain maps are used to construct a null distribution of the expected magnitude of model-empirical loading under random chance. Each sample in the null distribution (gray; N = 1000) was constructed by modulating gain in proportion to the values in a random spatial autocorrelation-preserving surrogate brain map, then re-computing model-empirical loading. Colored lines correspond to the different receptor-encoding genes (as reported in panel B). ** p < 0.01.