590 species and 152 genera in total

Cluster 1 (light blue): centered on Reserva Nacional Allpahuayo-Mishana, Peru (ER 737)
81 genera (63 unique), 271 species (234 unique), 184 samples, 4927 records
most common: Artibeus (138 samples), Artibeus lituratus (121 samples), Myotis (117 samples)

Cluster 2 (dark green): centered on Voinjama, Liberia (ER 716)
73 genera (33 unique), 292 species (187 unique), 75 samples, 1403 records
most common: Rhinolophus (48 samples), Hipposideros (43 samples), Myotis (35 samples)

Cluster 3 (red): centered on Badja (Session 2), Australia (ER 3066)
37 genera (8 unique), 117 species (40 unique), 26 samples, 428 records
most common: Chalinolobus (18 samples), Nyctophilus (17 samples), Vespadelus (17 samples)

Unclassified samples are in grey. These samples share no species with any of the cluster centers.

This cluster analysis tool is meant for educational purposes only. It uses a simplified and fast algorithm that it is related to the k-medoids method, which might yield better results. It first chooses the most diverse sample to be a cluster center (medoid), and then sequentially selects centers that are maximally dissimilar to those chosen already. It measures similarity using a variant of the Ochiai index that allows for non-zero values when no species are shared. After choosing the centers, it assigns the remaining samples to the clusters based on the same index. To speed up the computation, swaps are not considered.