Fig. 6

Modified pitch of ultrasonic vocalizations revealed by TrackUSF in Shank3-Het and KO. A 3D t-SNE analysis of all USFs recorded during all sessions. Each USF is represented by a dot, color-coded for the genotype and familiarity level. Black numbers represent the distinct clusters. Note the clear separation of cluster 1, which included non-vocal signals defined as noise. B Mean PSD profiles of all distinct clusters shown in B. The number of sessions represented by >10 USFs in each cluster, as well as the total number of USFs of each cluster, are detailed in the figure legend. Note the continuous spectrum in the 25–45-kHz range created by clusters 4–14. C Example spectrogram showing USFs from clusters 6–10, each marked as the number of the cluster it is associated with, superimposed by the TrackUSF software on their corresponding single USV. Note the gradual change in cluster number as the USV frequency is getting lower with time. D Example spectrograms showing USFs from several clusters, each marked as the number of the cluster it is associated with, superimposed by the TrackUSF software on their corresponding noise (i) or USVs (ii–v) signals. Note the trill-like appearance of USFs from cluster 16 (v). E Examples of vicinity curves describing the probability of a USF from any cluster (color-coded for the distinct clusters) to appear before or after USF from a given cluster across the three genotypes of Shank3-deficient rats. Note the stability across genotypes exemplified for clusters 9 (left) and 15 (middle), in contrast to the growing tendency of other clusters to combine with cluster 16 in Het and mainly KO animals. It should be noted that USFs of clusters 4–14 were so rare in WT animals that the vicinity curves of cluster 9 are taken from a single WT animal