Fig. 2
From: Systemic neurotransmitter responses to clinically approved and experimental neuropsychiatric drugs
Neurochemical similarity analysis. a Intra- and Inter-ATC code similarity of neurochemical response patterns across brain regions. Distributions are shown using kernel density estimation (KDE) with median similarities identified by the red marker and via the underlying boxplot with the notch denoting median, the outer boxes denoting the lower and upper-quartile ranges and the jitter scatter plot outlining the underlying observations. Compounds within the same ATC classification (intra-similarity) exhibit higher median similarity in neurochemical response than between ATC codes (inter-similarity), with a median response fingerprint Tanimoto coefficient’s of 0.43 and 0.33, respectively. The underlying distributions are statistically different with a two-sided Kolmogorov–Smirnov test p-value of 6.31e-56. However, the similarity distributions are quite wide (interquartile ranges span 0.0 to 1.0 for both intra- and inter-similarities), with large overlap. Hence, ATC codes could also be considered to be not very meaningful descriptors of neurotransmitter activities. b Intra-ATC class similarity of the neurochemical and chemical fingerprints of compounds for ATC classes with 4 or more representative compounds. The number of compounds underlying each distribution is denoted using (N = ), with the underlying distribution shown using KDE, the median shown via a solid red line and the underlying observations shown via the black stick markers. Our results highlight there are significant differences between different ATC classifications. N.B.: Only ATC classifications with N ≥ 4 (~3% of drugs in the database) are shown. The so-called combined subset category shows the distribution of the combined subset across the eight ATC classes shown and number of distinct compounds
