Figure 1.
Clustering procedure applied to subset of compendium. To reduce the contributions of highly replicated conditions in the compendium, we applied agglomerative hierarchical clustering using a distance metric based on Spearman’s rank correlation to the full compendium, then selected and consolidated clusters with a maximum intra-cluster distance of 0.3 and a minimum size of 5. We selected 323 neuron samples that clustered together to graphically demonstrate this preprocessing step. Experiments from the original compendium (left) clustered together to form meta-samples (right) with corresponding colors. We excluded samples with black lines (left) from the meta-compendium because they did not satisfy our cluster selection criteria: they might represent extreme experimental perturbations or technical errors.