Predicting cochlear synaptopathy in mice with varying degrees of outer hair cell dysfunction using auditory evoked potentials

An example Jupyter notebook is included to demonstrate how to generate predictions using the cross-validation approach described in the companion work. The notebook is also available to view in HTML format if Jupyter is not available. Data includes: Synapse counts for all frequencies (synapses.csv) DPOAE and ABR thresholds for all frequencies (thresholds.csv) The DPOAE, ABR, and EFR measures used in the synapse prediction algorithms for 16 and 32 kHz (measures.csv) The harmonics used for the synapse prediction algorithms in supplemental data figure 1 (RAM_1000_harmonics.csv) The synapse predictions (cv_predictions.csv) The synapse predictions generated from the harmonics used for supplemental data figure 1 (cv_predictions_RAM_1000_harmonics.csv). In measures.csv, columns are as follows: tx - Rough categorization of treatment group (e.g., young, acute noise-exposed, etc.). animal_id - Unique subject identifier that can be used to align spreadsheets (all data is from the left ear only). ear - Ear of subject. frequency - Stimulus or cochlear frequency (e.g., EFR carrier frequency, ABR tone pip frequency, DPOAE f2 frequency, cochlear frequency for histology). dp_th - DPOAE threshold in dB SPL. dp_level_40 - DPOAE level at L2 = 40 dB SPL. dp_level_55 - DPOAE level at L2 = 55 dB SPL. synapses - Average number of synapses per inner hair cell. CtBP2 - Average number of CtBP2 puncta per inner hair cell. The difference between synapses and CtBP2 is the number of orphan ribbons. sex - Sex of mouse. ABR-w1-40 - ABR wave 1 amplitude at 40 dB SPL in uV. ABR-w1-50 - ABR wave 1 amplitude at 50 dB SPL in uV. ABR-w1-60 - ABR wave 1 amplitude at 60 dB SPL in uV. ABR-w1-70 - ABR wave 1 amplitude at 70 dB SPL in uV. ABR-w1-80 - ABR wave 1 amplitude at 80 dB SPL in uV. w1-slope-all - ABR wave 1 slope in uV/dB for all supra-threshold levels. w1-slope-4 - ABR wave 1 slope in uV/dB for the four highest stimulus levels. w1-slope-2 - ABR wave 1 slope in uV/dB for the two highest stimulus levels. RAM-110 - RAM EFR magnitude in dB re 1uV for the sum of f0 ... f4 for fm = 110 Hz. RAM-110_nf-110 - RAM EFR magnitude in dB SNR for the sum of f0 ... f4 for fm = 110 Hz. RAM-110_f0-110 - RAM EFR magnitude in dB re 1 uV for f0 only for fm = 110 Hz. RAM-110_f0_nf-110 - RAM EFR magnitude in dB SNR for f0 only for fm = 110 Hz. RAM-PLV-110 - RAM EFR phase locking value for f0 for fm = 110 Hz. Columns are similarly represented for RAM with a modulation frequency of 1000 Hz, SAM with a modulation frequency of 110 Hz, and SAM with a modulation frequency of 1000 Hz. Missing data are indicated by blank cells. For thresholds.csv: DPOAE - DPOAE threshold in dB SPL. ABRpresto - ABR threshold in dB SPL as computed by ABRpresto (https://doi.org/10.1016/j.heares.2025.109258). Other columns are represented as in measures.csv. For RAM_1000_harmonics.csv, the individual harmonics are in units of dB re 1 uV and represented as f0 through f4. Other columns are represented as in measures.csv. For synapses.csv, columns are represented as in measures.csv. For cv_predictions.csv and cb_predictions_RAM_1000_harmonics.csv: measure - The measures incorporated into the model (blank means no measure). dp - The DPOAE correction used (blank means no DPOAE correction). synapses - The actual (observed) synapse count. pred_synapses - The prediction of the model for that particular repeat and fold. repeat - Repeat number for cross validation. fold - Fold number for cross validation. For the RAM 1000 predictions, the measure column indicates the model name used for the prediction shown in supplemental data figure 1. RAM maps to f₀₋₄ and RAM_power maps to f²₀₋₄. For all other combinations of harmonics, f0_1 maps to f₀ · f₁ with other combinations represented similarly. This data is made available under Creative Commons Attribution 4.0 International. Changelog: Version 1 - Initial upload of data at time of manuscript submission. Version 2 - Deleted. Uploaded new file (see Version 3), but did not realize the existing files in Version 1 had to be re-uploaded as well. Deleted in favor of Version 3, which is a comprehensive set of all files. Version 3 - Added cv_predictions_RAM_1000.csv in response to reviewer feedback.