custom scripts and pre-programmed gui functions in matlab r2020a 9.8.0  (MathWorks Inc)

Journal: bioRxiv

Article Title: Paternal aging affects developmental convergence of vocal behavior in individual mice: machine learning-driven analyses of individuality

doi: 10.1101/738781

Figure Lengend Snippet: (A) The experimental design in the model of paternal aging. (B) Typical sonograms of twelve types of syllables. Scale bar = 60 kHz, 20 ms. (C) Following maternal separation, USVs were recorded from individual pups one by one. The resulting USVs were processed to spectrogram by USVSEG, which segmented the syllables and outputted the syllable images with acoustic parameters. Subsequently, the VAE was trained to learn and encode the syllables. The resulting syllables were visualized and mapped in a latent space.

Article Snippet: Acoustic waveforms were first processed using a GUI-based MATLAB script (“USVSEG”) originally developed for segmenting USVs emitted by rodents ( ) because we found that an audio segmentation method implemented in the original variational autoencoder (VAE) detected large amounts of noise.

Techniques:

Journal: bioRxiv

Article Title: Paternal aging affects developmental convergence of vocal behavior in individual mice: machine learning-driven analyses of individuality

doi: 10.1101/738781

Figure Lengend Snippet: (A) Z scores of syllable number, duration, maximum frequency, maximum amplitude, and interval. The thick horizontal lines indicate the mean value of Z scores. Boxplots represent the range of Z scores. The horizontal lines in the boxes signify respective group medians, boxes are 25 th – 75 th percentiles, and whiskers represent minimum to maximum. The light blue point depicts the outlier. Each point represents the syllable data of one pup. * p < 0.05, ** p < 0.01, *** p < 0.001 indicate a significant difference between the wildtype and Auts2 del8/+ pups (Wilcoxon test). (B) Pie graphs show the percentage composition of the twelve types of syllables in the pups of wildtype and Auts2 del8/+ . The colors indicate the syllable type shown in . (C) A significant decrease of the number of syllable types in Auts2 del8/+ pups. The thick horizontal lines indicate the mean value of the number of syllable types. Boxplots represent the range of the number of syllable types. The horizontal lines in the boxes signify respective group medians, boxes are 25 th – 75 th percentiles, and whiskers represent minimum to maximum. Each point represents the data of one pup. *** p < 0.001 indicates a significant difference (Wilcoxon test). (D) A significant decrease of entropy scores of syllables in Auts2 del8/+ pups. The thick horizontal lines indicate the mean value of entropy scores. Boxplots represent the range of entropy scores. The horizontal lines in the boxes signify respective group medians, boxes are 25 th – 75 th percentiles, and whiskers represent minimum to maximum. Each point represents the data of one pup. *** p < 0.001 indicates a significant difference (Wilcoxon test). (E) The VAE takes spectrograms from the syllables of wildtype and Auts2 del8/+ pups as input and visualizes them as a vector of latent dimensions at P7. Each point in this latent space represents a single spectrogram of a syllable. The distance between points indicates neighbor similarity. (F) Based on the USVSEG, PCA summarized the patterns of individual wildtype and Auts2 del8/+ pups. Each point represents the syllable pattern of one pup. Shorter distances between points indicate higher similarity in syllable patterns. Grey and light blue clusters contain 90% of the population of wildtype and Auts2 del8/+ pups, respectively. (G) Based on the VAE, t-SNE summarized the patterns of individual wildtype and Auts2 del8/+ pups. Each point represents the syllable patterns of one pup. Shorter distances between points indicate higher similarity in syllable patterns. Grey and light blue clusters contain 90% of the population of wildtype and Auts2 del8/+ pups, respectively.

Article Snippet: Acoustic waveforms were first processed using a GUI-based MATLAB script (“USVSEG”) originally developed for segmenting USVs emitted by rodents ( ) because we found that an audio segmentation method implemented in the original variational autoencoder (VAE) detected large amounts of noise.

Techniques: Plasmid Preparation

Journal: bioRxiv

Article Title: Paternal aging affects developmental convergence of vocal behavior in individual mice: machine learning-driven analyses of individuality

doi: 10.1101/738781

Figure Lengend Snippet: (A) Based on the USVSEG, PCA summarized the patterns of individual YFO and AFO at P3, P6, P9, and P12. Each point represents the syllable pattern of one pup. Shorter distances between points indicate higher similarity in syllable patterns. Grey and light red clusters contain 90% of the population of the YFO and AFO, respectively. The M and F mean male and female, respectively. (B) Based on the VAE, t-SNE summarized the patterns of individual YFO and AFO at P3, P6, P9, and P12. Each point represents the syllable features of one pup. Shorter distances between points indicate higher similarity in vocal repertoire. Grey and light red clusters contain 90% of the population of the YFO and AFO, respectively. The M and F mean male and female, respectively. (C) Summary of individual traces at P3 and P12 based on (A) and (B). Each point represents the syllable pattern of one pup.

Article Snippet: Acoustic waveforms were first processed using a GUI-based MATLAB script (“USVSEG”) originally developed for segmenting USVs emitted by rodents ( ) because we found that an audio segmentation method implemented in the original variational autoencoder (VAE) detected large amounts of noise.

Techniques:

Journal: bioRxiv

Article Title: Paternal aging affects developmental convergence of vocal behavior in individual mice: machine learning-driven analyses of individuality

doi: 10.1101/738781

Figure Lengend Snippet:

Article Snippet: Acoustic waveforms were first processed using a GUI-based MATLAB script (“USVSEG”) originally developed for segmenting USVs emitted by rodents ( ) because we found that an audio segmentation method implemented in the original variational autoencoder (VAE) detected large amounts of noise.

Techniques: