Data from: Vocal repertoire expansion in singing mice by co-opting a conserved midbrain circuit node
Data files
Nov 26, 2025 version files 2.33 GB
-
heliox.zip
116.52 KB
-
pag_opto.zip
117.34 KB
-
pag_telc.zip
9.81 MB
-
README.md
24.44 KB
-
solo_social.zip
1.45 GB
-
thermistor.zip
869.62 MB
Abstract
How neural circuits generate diverse behaviors is a fundamental question in neuroscience. Distinct behavioral outputs may arise from either dedicated motor circuits or shared circuits operating in different functional states. While multifunctional circuits offer an efficient solution for behavioral flexibility and may drive rapid evolutionary adaptations, their neural mechanisms remain poorly understood, especially in mammals. Here, we leverage the rich vocal repertoire of the singing mouse (Scotinomys teguina) to investigate the organizational logic of multifunctional motor circuits. We developed a behavioral assay (PAIRId) that enables precise attribution of vocalizations during social interactions. This paradigm revealed two distinct vocal modes: soft, variable, ultrasonic vocalizations (USVs) ancestral to rodents used for short-range communication, and loud, rhythmic, human-audible songs unique to the singing mouse lineage used for long-range communication. Despite their substantial acoustic and contextual differences, we found that USVs and songs do not arise from parallel pathways. Instead, they share the same sound production mechanism, phonatory-respiratory coupling, and vocal gating from the midbrain caudolateral periaqueductal gray (clPAG). To understand the mechanism governing song production, we combined mathematical modeling of song rhythm with synaptic silencing of clPAG, which progressively reduced song amplitude and duration. We demonstrate that song duration decreases via a single parameter controlling its termination. Notably, this mechanism also accounts for sexual dimorphism in songs, identifying clPAG as a key locus for driving natural behavioral variability. Our findings reveal how parametric tuning of a central circuit node produces distinct vocal modes, providing a mechanistic basis for rapid behavioral evolution in mammals.
Dataset DOI: 10.5061/dryad.ht76hdrw5
Description of data and file structure
This dataset comprises all data required to reproduce the quantitative analyses and figures presented in the associated manuscript when processed using the accompanying analysis code. Data are organized into .zip archives corresponding to individual experiments. Each archive contains a metadata .csv file and one or more .csv files generated from audio analyses, which include note onset and offset times as well as computed acoustic features. Experiments that incorporate modalities beyond audio additionally include the corresponding processed data files.
Dataset contents
solo_social.zip
Data from intact adult singing mice male–female close-range social interactions within the PAIRId paradigm. Two cohorts (each containing three males and two females) were recorded across ten sessions, each lasting five hours, encompassing all opposite-sex dyads (60 hours total).
Included files comprise processed audio vocalization timestamps from the PAIRId assignment algorithm and manual curation, as well as tracked positions of both mice throughout each session in common world coordinates.
The .zip contains two metadata files (solosocial_sessions.csv, social_sessions_samples_info.csv) and three main data directories:
08_solosocial/,09_solosocial_02/— Scotinomys teguina recordings from solo and social contexts.mmus/— Mus musculus (laboratory mouse) dyad recordingscam_params/— YAML files containing camera calibration and rig configuration parameters for video alignment and world coordinate reconstruction.
Each Scotinomys session folder is named according to the format
sessionNumber_date_maleID_femaleID_condition (e.g., 03_20230921_um006_uf004_mf) and includes two subcontexts:
01_solo/— recordings from individual animals in acoustic isolated boxes, labelledtopandbottom.02_social/— recordings from when the mice were allowed to interact within PAIRId, with synchronized video tracking (process_video)and multi-microphone audio detections (process_audio)
File descriptions
-
solosocial_sessions.csv— Metadata linking each recording session to the corresponding subjects.Column Description sessionUnique session identifier (date + subject IDs) basepathFilepath mouse_1,mouse_2Identifiers of the recorded individuals (male and female, respectively) -
social_sessions_samples_info.csv— Summary of total recording frame counts per session, used to relate acoustic and video sample indicesColumn Description Sequential index session_idSession identifier matching entries in solosocial_sessions.csvaudio_total_framesTotal number of frames in the processed audio data video_total_framesTotal number of frames in the synchronized video data -
all_notes_corrected_features.csv— Note-level acoustic feature tables located within./process_audio/all_notesthen the audio file number folder (0000###/).Column Description # Sequential note index start,endStart and end time of the note within the session (seconds) durationDuration of the note (seconds) maxfreq,minPlaceholder numeric values displayed in the detection GUI; these do not represent measured frequencies and are not biologically meaningful typeAutomatically assigned call type label (not curated) codeFor social data recorded in PAIRId, indication of assigned source ( 0= left,1= right,6= unknown). This column missing and not applicable for solo data.max_amp_high,max_amp_songPeak amplitude of the note within the 10–120 kHz frequency range as calculated on high-gain audio ( high) or low-gain audio (song)note_rateInstantaneous note rate, calculated as the reciprocal of the interval between the start times of two consecutive notes (1 / Δstart). Reflects the local tempo of vocal production. isiIntervening silence interval, defined as the time difference between one note’s end and the next note’s start. Represents the silent gap between consecutive notes. -
common_coords_00###.npz— mouse positions estimated using a SLEAP model (stored within./process_video/common_coords. The final 5 digit number of each filename indicates the video index.Variable Description l_coordinates_world,r_coordinates_worldFor every video frame, world-coordinate (x, y) positions for the six SLEAP pose-estimation nodes for each mouse ( l_= left mouse,r_= right mouse). Nodes include, in order: right ear, left ear, nose, mid-head, mid-back, and tail base.
heliox.zip
Data from four unique singing mice male–female dyads recorded while vocalizing first in normal atmospheric conditions and subsequently with heliox gas introduced into the enclosure. Included files contain the fundamental frequency measurements of selected note classes for each pair, encompassing ultrasonic vocalizations (USVs), song notes, and squeaks
File descriptions
-
heliox_sessions.csv— metadata table summarizing each recording session.Column Description sessionUnique session identifier dateRecording date male_id, female_idSubject identifiers for each dyad atm_num_song_notes_quantifiedNumber of song notes analyzed in air atm_num_usvs_quantifiedNumber of ultrasonic vocalizations analyzed in air atm_num_squeaks_quantifiedNumber of squeaks analyzed in air hel_num_song_notes_quantifiedNumber of song notes analyzed in heliox hel_num_usvs_quantifiedNumber of ultrasonic vocalizations analyzed in heliox hel_num_squeaks_quantifiedNumber of squeaks analyzed in heliox -
helium_manual_calls.csv,helium_manual_songs.csv,helium_manual_squaks.csv— Quantified acoustic features for vocalization modes.Column Description trialUnique identifier for the recording trial pathFile path file_callName of source audio file containing the call file_detName of source detection file containing the call airRecording atmosphere ( atm= air,hel= heliox)classVocalization class ( downsweeportrill= USV subtype,song= song note, orsqueak)callStart,callStopOnset and offset time of the call within the file (seconds) f0_hi,f0_loHighest and lowest measured portion of the fundamental frequency (Hz)
thermistor.zip
Data from a male singing mouse implanted with an intranasal thermistor used as a proxy for respiration. Included files contain detected vocalization timestamps recorded when the focal male was either alone or paired with a mute female, along with the corresponding analog thermistor signals acquired using an Intan recording controller.
The .zip contains a metadata file (thermistor_sessions.csv) and three session folders named according to the format
YYMMDD_mouseID_sessionNumber. Each session folder includes two components:
process_audio/— processed acoustic detections and note-level acoustic features.process_intan/— synchronized thermistor voltage traces recorded via an Intan acquisition system (.matformat).
File descriptions
-
thermistor_sessions.csv— Metadata linking each session to the recording base path and subject ID.Column Description sessionUnique session identifier (date + ID code) basepathRelative file path for the session folder mouse_idIdentifier of the implanted subject therm_clippedLogical indicator for whether thermistor signal contained clipping artifacts ( TRUE/FALSE) -
all_notes_corrected_features.csv— Per-note acoustic feature table generated from processed audio.Column Description #Sequential note index start,endStart and end time of the note with the session (seconds) durationDuration of note (seconds) maxfeq,minPlaceholder numeric values displayed in the detection GUI; these do not represent measured frequencies and are not biologically meaningful typeAutomatically assigned call type label (not curated) max_amp_highPeak amplitude of the note within the 10–120 kHz frequency range -
therm_*_*.mat-- Each.matfile inprocess_intan/contains synchronized analog recordings acquired using an Intan system. Files are named according to the session date and start time (e.g.,therm_231017_152230.mat) and include four variables:Variable Description cam_trigs_expDigital trigger channel indicating camera frame times used as synchronization events. ch1_audio_expAudio channel 1 waveform recorded from one microphone input ch2_audio_expContinuous analog voltage trace from the intranasal thermistor sensor, used as a proxy for respiratory airflow. This is the primary signal of interest for respiration–vocal coupling analyses. d2_expPlaceholder variable representing an unused digital input channel in this dataset (values are all 0).
pag_opto.zip
Data from four male singing mice subjected to unilateral optogenetic stimulation of the caudolateral periaqueductal gray (clPAG) using excitatory opsins. Included files contain detected vocalization timestamps along with corresponding timestamps for optogenetic stimuli of one, two, or four seconds in duration.
The .zip contains a metadata file (pag_opto_usv_sessions.csv) and multiple session folders named according to the format
YYMMDD_mouseID_sessionNumber. Each session folder includes two components:
process_audio/— processed acoustic detections and note-level acoustic features.process_intan/— synchronized optogenetic stimulation timestamps (.csvformat).
File descriptions
-
pag_opto_usv_sessions.csv— Metadata summarizing all optogenetic recording sessions.Column Description sessionUnique session identifier (date + ID code) AnimalSubject identifier DurationDuration of optogenetic stimuli used in the session (1, 2, or 4 seconds) h_audio_file_sraudio file sampling rate h_audio_real_srestimated audio sampling rate for synchronization -
h_T*_corrected.csv— Per-note acoustic feature table generated from processed audio.Column Description #Sequential note index start,endStart and end time of the note within the session (seconds) durationDuration of the note (seconds) maxfreq,minPlaceholder numeric values displayed in the detection GUI; these do not represent measured frequencies and are not biologically meaningful typeAutomatically assigned call type label (not curated) max_amp_low,max_amp_highPeak amplitude of the note within the 10–120 kHz frequency range as calculated on low-gain audio ( low) or high-gain audio (high) -
opto_stims.csv— Optogenetic stimulation timestamps.Column Description stim_starts,stim_stopsStart and end times of optogenetic stimulus (seconds)
pag_telc.zip
Data from five male singing mice subjected to bilateral synaptic silencing of neurons in the caudolateral periaqueductal gray (clPAG) via expression of tetanus toxin light chain (TeLC). Included files comprise processed audio vocalization timestamps generated by the PAIRId assignment algorithm and manual curation from recordings collected the day before surgery and during the five to six days following surgery, as the mice produced progressively deteriorating vocalizations before becoming mute.
The .zip contains metadata files (pag_telc_sessions.csv, pag_telc_curation.csv) and multiple session folders named according to the format
YYMMDD_mouseID_sessionNumber. Each session folder includes one or more subdirectories containing the processed acoustic data:
process_audio/ororiginal_audio/— folders containing extracted note-level acoustic features inall_notes_corrected_features.csvformat, organized by audio file number (0000###).
File descriptions
-
pag_telc_sessions.csv— Metadata linking each session to the recording path, experimental condition, and post-surgery time point.Column Description sessionUnique session identifier (date + subject IDs) basepathFilepath mouse_1,mouse_2Identifiers of the recorded individuals conditionExperimental condition ( pre= pre-surgery,post= post-surgery)processeddate on which the data were processed ( YYMMDD)anchor_file_idxthe initial audio file index num_post_daysnumber of days recorded num_vocal_daysNumber of days in which vocalizations were detected -
pag_telc_curation.csv— Manual curation log corresponding to PAIRId output and post-processing review.Condition Description exptitleExperiment title sessionSession identifier basepathRelative file path for the session folder mouse_1,mouse_2Identifiers of the recorded individuals conditionExperimental condition ( pre= pre-surgery,post= post-surgery)idxsequential position of curated file hraudio file number corresponding to curated file lights_turn_onLogical indicator for whether this was a file where the light turned on for the mice ( 1=TRUE /0= FALSE)truepos_usvsCount of manually verified ultrasonic vocalizations (true positives); provided as partial internal annotation notes rather than a comprehensive count. -
all_notes_corrected_features.csv— Per-note acoustic feature table generated from processed and manually curated recordings.Column Description Sequential note index start,endStart and end time of the note within the session (seconds) durationDuration of the note (seconds) maxfreq,minPlaceholder numeric values displayed in the detection GUI; these do not represent measured frequencies and are not biologically meaningful typeAutomatically assigned call type label (not curated) codeIndication of assigned source ( 0= left,1= right,6= unknown)max_amp_high,max_amp_songPeak amplitude of the note within the 10–120 kHz frequency range as calculated on high-gain audio ( high) or low-gain audio (song)
Code/software
The code necessary to reproduce the results in the associated publication are available on GitHub.
https://github.com/singingmicelab/zheng-harpole-2025-vocalmodes-pag
