DANDI:000931

⊘Blocked37

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 1; Figure 1C; STAR Methods, Inhalation clustering· Data missing from the dataset

  Inhalation durations appear bimodal, and a Gaussian-mixture and hierarchical clustering of inhalation waveforms sorts them into slow and fast inhalations; fast inhalations are shorter and have steeper airflow changes than slow ones.

  Since the distribution of inhalation durations appeared bimodal, we developed a semi-unsupervised method that sorted inhalations into slow and fast depending on their airflow rate waveforms ... Fast inhalations had shorter duration and steeper changes in airflow rate than slow inhalations.

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 1; Figure 1C; STAR Methods, Inhalation clustering

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, which the criterion needs; it also pools the eight awake mice of the leave-one-mouse-out test, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figures 1E and S1A· Data missing from the dataset

  PCx population activity encodes the inhalation phase; a cross-validated support vector regression predicts the phase above a shuffle null in individual mice.

  Hence, PCx activity encoded the phase and speed of each inhalation (Figures 1E and 1F; Figures S1A and S1B).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figures 1E and S1A

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times or the brain area of each unit, which the criterion needs; it also pools the awake mice whose phase decoding the paper reports, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedFigure S1B; Results, Sniffing causes unstable odor responses in the PCx, paragraph 3· Data missing from the dataset

  The inhalation speed (fast against slow) can be decoded above chance from the mean firing rate in the first 180 ms of inhalation in individual mice.

  Decoding the inhalation speed (fast versus slow) using the mean firing rate in the early 180 ms of inhalation in individual mice.

  Figure S1B; Results, Sniffing causes unstable odor responses in the PCx, paragraph 3

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs. The paper says its processed data are available on request, and no public source holds them. The criterion asks for every mouse, and the other five mice of Figure 1 are not in the dandiset; even the one mouse in the file cannot be tested, as its inhalations are not in it.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure 1H legend· Data missing from the dataset

  Of 660 PCx neurons in six mice, 31.1% (205) reliably respond to at least one odor concentration.

  Of the recorded neurons, 31.1% reliably responded to at least one odor concentration (205 of 660 neurons; six mice).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure 1H legend

  The dandiset's one file holds spike times of one mouse and none of the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figures 1H and 1I· Data missing from the dataset

  In 72.2% of odor-responsive neurons (148 of 205) the response amplitude during inhalation depends on how fast the animal inhales, more than expected by chance.

  Strikingly, how fast an animal inhaled influenced the response amplitude during inhalation in 72.2% of those neurons (148 of 205 neurons; Figures 1G–1I).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figures 1H and 1I

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure 1H· Data missing from the dataset

  A faster inhalation increases 37.3% of odor responses and decreases 19.2% of them.

  The impact of a faster inhalation was not uniform across odor responses: 37.3% of the odor responses increased with a rapid inhalation, whereas 19.2% decreased in amplitude.

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure 1H

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure S1C· Data missing from the dataset

  The sniff-driven variability in odor responses is already significant in the first 70 ms of an inhalation.

  This sniffdriven variability in odor responses was already significant during the first 70 ms of an inhalation (Figure S1C).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure S1C

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure S1D· Data missing from the dataset

  Regular-spiking (putatively excitatory or feedforward inhibitory) and fast-spiking (putatively feedback inhibitory) neurons show similar effects of inhalation speed on their odor responses.

  After sorting neurons into regular spiking (RS; putatively excitatory or feedforward inhibitory) and fast spiking (FS; putatively feedback inhibitory), we observed similar effects of the inhalation speed in both types (Figure S1D).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 3; Figure S1D

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration, unit waveforms or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 4; Figure 1J· Data missing from the dataset

  Across odor-responsive neurons, the absolute response change caused by a slow-to-fast switch in inhalation (at 0.01%) is larger than that caused by a tenfold increase in concentration (0.01 to 0.1%, slow inhalations).

  On average, sniffs caused a response amplitude change bigger than that caused by a 10-fold increase in odor concentration in the odor-responsive neurons (Figure 1J).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 4; Figure 1J

  What the paper reports

  • Wilcoxon signed-rank test, p < 10⁻⁴ (n = 477 neuron-odor pairs of odor-responsive neurons)

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 4; Figure 1L· Data missing from the dataset

  Concentration discriminability (CDI) of single neurons is lower when slow and fast inhalations are mixed than when computed within one inhalation speed.

  Combining responses from slow and fast inhalations decreased the CDI compared with responses observed during uniform inhalation speeds (Figure 1L).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 4; Figure 1L

  What the paper reports

  • Wilcoxon signed-rank test, p < 10⁻⁴ (n = 477 neuron-odor pairs)

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other five of the six mice of Figure 1, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 2· Data missing from the dataset

  A regularised Poisson GLM with inhalation speed, concentration and their interaction explains 91.9% of the variance in the mean spike count during the first 180 ms of odor inhalation across neurons.

  This model explained 91.9% of the variance in the mean number of spikes generated during the first 180 ms of odor inhalation across all neurons.

  Results, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 2

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure 2C· Data missing from the dataset

  In the GLM, the interaction between inhalation speed and concentration has a small effect on the odor response, while the inhalation-speed (non-olfactory) term has a larger additive or subtractive effect.

  The GLM analysis revealed that the interaction between inhalation speed and concentration response had a small effect on the odor response amplitude during an inhalation (Figure 2C). Instead, the non-olfactory input had a more significant additive or subtractive effect (Figure 2C).

  Results, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure 2C

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure 2D· Data missing from the dataset

  The signs of the inhalation-speed and concentration weights can differ within a neuron, so that a fast inhalation can raise or lower a response regardless of how the neuron responds to a higher concentration; the combination is heterogeneous across neurons.

  Importantly, the sign of the weight for the non-olfactory input and the odor concentration factors could differ in the same neuron (Figure 2D). The non-olfactory input caused by a fast inhalation could increase or decrease the odor response amplitude regardless of how the neuron responded to a higher concentration of that odor.

  Results, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure 2D

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figures S2A and S2B· Data missing from the dataset

  RS and FS neurons both show a heterogeneous distribution of inhalation-speed and concentration sensitivities.

  We similarly found a heterogeneous distribution of non-olfactory sensitivities in RS and FS neurons (Figures S2A and S2B).

  Results, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figures S2A and S2B

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration, unit waveforms or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure S2C; STAR Methods, Prediction of encoding properties from waveforms· Data missing from the dataset

  Spike waveform features (asymmetry and trough-to-peak delay) cannot predict a neuron's inhalation-speed weight, concentration weight, or both, with any of three models.

  In none of the 9 cases, the predictive model could extract a significant relationship (as measured by the cross-validated R2 of the regression) between the spike waveform and the encoding parameters.

  Results, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure S2C; STAR Methods, Prediction of encoding properties from waveforms

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration, unit waveforms or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure S2D· Data missing from the dataset

  Restricting the GLM analysis to the first 70 ms of inhalation gives similar results.

  Finally, restricting the analysis to the early 70 ms of inhalation also gave similar results (Figure S2D).

  Results, PCx neurons randomly mix olfactory and non-olfactory inputs during inhalations, paragraph 3; Figure S2D

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 2; Figure 3C· Data missing from the dataset

  In tracheostomized anesthetized mice, 57.1% of PCx neurons (144 of 252) encode the rate of artificial nasal airflow, 117 with positive and 27 with negative slopes.

  We found that 57.1% of neurons displayed mechanosensory responses, meaning that they encoded the airflow rate.

  Results, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 2; Figure 3C

  The dandiset's one file holds spike times of one mouse and none of the times and rates of the artificial suction pulses or the brain area of each unit, which the criterion needs; it also pools the three tracheostomized anesthetized mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 2; Figures 3B and S3B· Data missing from the dataset

  Faster artificial airflow shortens response latency without significantly changing the order in which neurons are activated.

  Faster airflow also decreased the response latency without significantly changing the sequential activation order across neurons (Figure 3B; Figure S3B).

  Results, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 2; Figures 3B and S3B

  The dandiset's one file holds spike times of one mouse and none of the times and rates of the artificial suction pulses or the brain area of each unit, which the criterion needs; it also pools the three tracheostomized anesthetized mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 3; Figures 3E and 3F· Data missing from the dataset

  In awake, normally breathing mice without odor, 78.3% of PCx neurons (748 of 955, 10 mice) and 77.2% of odor-responsive neurons encode the peak inhalation airflow with a monotonic change; 36.7% increase and 41.7% decrease their response with faster inhalation.

  This analysis revealed that 78.3% of the PCx neurons (77.2% of the odorresponsive neurons) encoded the peak inhalation airflow rate with a monotonic change in spiking activity after the inhalation (Figure 3E) ... 36.7% of the neurons increased the amplitude of their responses when the inhalation speed increased, and another 41.7% decreased their responses

  Results, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 3; Figures 3E and 3F

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the ten control mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 4; Figures S3D and S3E· Data missing from the dataset

  Cutting the anterior ethmoidal nerve leaves PCx responses to inhalation intact; airflow tuning curves after neurectomy (551 of 752 neurons tuned, 5 mice) are similar to those of control mice.

  Severing the AEN left intact the responses of PCx neurons to inhalations. The tuning curves for the airflow rate after the neurectomy remained similar to those of control mice (Figures S3D and S3E).

  Results, A mechanosensory input signals the nasal airflow rate to the PCx, paragraph 4; Figures S3D and S3E

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times or the brain area of each unit, which the criterion needs; it also pools the five neurectomized mice and the ten control mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figures S4A and S4B· Data missing from the dataset

  After bulbectomy, PCx neurons do not respond to odors and the population does not encode the inhalation phase.

  In bulbectomized mice, PCx neurons were unresponsive to odors (Figure S4A). Moreover, the population activity did not encode the inhalation phase (Figure 4B; Figure S4B).

  Results, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figures S4A and S4B

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the four bulbectomized mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4D· Data missing from the dataset

  After bulbectomy, 15.7% of PCx neurons still fire more during fast sniffs than normal inhalations, against 29.7% in control mice.

  Nonetheless, 15.7% of PCx neurons exhibited increased spiking activity during fast sniffs compared with normal inhalations (see example neuron in Figure 4C). For comparison, this proportion was 29.7% in control mice.

  Results, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4D

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the four bulbectomized and ten control mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4E· Data missing from the dataset

  In bulbectomized mice the sniff-induced response of sniff-preferring PCx neurons is smaller in amplitude than in control mice.

  However, the sniff-induced activity in bulbectomized mice was smaller in amplitude and more variable across sniffs than in control mice (Figures 4D–4F).

  Results, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4E

  What the paper reports

  • two-tailed t test with unequal variances, p = 2 × 10⁻⁴ (n = 284 control and 23 bulbectomized sniff-preferring neurons)
  • PCx control, PCx bulbectomized and MCx control (Figure S4D): 1-way ANOVA, Tukey-Kramer post hoc, p < 10⁻⁴

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the four bulbectomized and ten control mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedFigure 4F; Results, A top-down input may supply behavioral state information during a sniff, paragraph 2· Data missing from the dataset

  The coefficient of variation of the response across first sniffs is larger in bulbectomized than in control mice.

  Coefficient of variation of the response amplitude across first sniffs in control and bulbectomized mice (p < 0.0001; two-tailed t test).

  Figure 4F; Results, A top-down input may supply behavioral state information during a sniff, paragraph 2

  What the paper reports

  • two-tailed t test, p < 10⁻⁴ (n = sniff-preferring neurons, 284 control and 23 bulbectomized)
  • PCx control, PCx bulbectomized and MCx control (Figure S4E): 1-way ANOVA, Tukey-Kramer post hoc, p < 10⁻⁴

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the four bulbectomized and ten control mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4G· Data missing from the dataset

  A linear classifier distinguishes sniffs from regular inhalations from PCx activity in both control and bulbectomized mice, but with clearly lower accuracy after bulbectomy.

  Correspondingly, while a linear classifier could differentiate a sniff from a regular inhalation based on PCx activity in both groups, its accuracy was notably reduced post-bulbectomy (Figure 4G).

  Results, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4G

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the four bulbectomized and ten control mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4D· Data missing from the dataset

  In bulbectomized mice the sniff-induced activity starts rising several tens of milliseconds before the nominal sniff onset.

  Remarkably, the sniff-induced activity observed in the bulbectomized mice seemed to start rising several tens of milliseconds before the nominal onset of the sniff.

  Results, A top-down input may supply behavioral state information during a sniff, paragraph 2; Figure 4D

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the four bulbectomized mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, A top-down input may supply behavioral state information during a sniff, paragraph 3; Figure S4C· Data missing from the dataset

  A similar peri-sniff increase occurs in motor cortex, where a slow inhalation modulates 5.8% of neurons but 17.1% show a significant peri-sniff increase.

  Here, a slow inhalation modulated the activity of 5.8% of neurons (see STAR Methods). In contrast, 17.1% of MCx neurons showed a significant peri-sniff increase in spiking activity

  Results, A top-down input may supply behavioral state information during a sniff, paragraph 3; Figure S4C

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the mice with motor cortex units, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Odor information and inhalation information are independent, paragraphs 2 and 3; Figures 5B, S5C and S5E· Data missing from the dataset

  Inhalation speed is decoded with high accuracy from the first PC of the population odor response but concentration is not; concentration (and odor identity) are decoded from the second and third PCs, from which speed is not.

  Linear decoders—namely, linear support vector machine (lSVM) classifiers—could discriminate the inhalation speed with high accuracy using the projection of neural activity onto the first PC. In contrast, decoders could not classify an odor concentration using only the activity explained by the first PC.

  Results, Odor information and inhalation information are independent, paragraphs 2 and 3; Figures 5B, S5C and S5E

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice and the four odor-identity mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Odor information and inhalation information are independent, paragraph 4; Figure 5C· Data missing from the dataset

  The population directions that separate concentrations during slow and during fast inhalation differ but are more similar to each other than to the direction that separates inhalation speeds.

  We found that, even though these two directions were different (as expected, given the high dimensionality of the population response), they were more similar than the directions that tell how fast a mouse breathes in apart from how concentrated the odor is (Figure 5C).

  Results, Odor information and inhalation information are independent, paragraph 4; Figure 5C

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, The distributed representation of odor concentration is sniff invariant, paragraph 1; Figures 5D and S5F· Data missing from the dataset

  Concentration decoders trained on one inhalation speed generalize to the other (trans-decoding) with performance similar to cis-decoding, both on PCs 2 and 3 and on the full population, so the PCx representation of concentration is sniff invariant.

  A decoder using PCx response projections on the second and third PCs exhibited similar cis- and trans-decoding performances, suggesting that PCx has sniffinvariant representations of odor concentration (Figure 5D; Figures S5F and S5G).

  Results, The distributed representation of odor concentration is sniff invariant, paragraph 1; Figures 5D and S5F

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, The distributed representation of odor concentration is sniff invariant, paragraph 1; Figures S5G and S5H· Data missing from the dataset

  Cis- and trans-decoding results are similar with spike counts over 70 ms or 180 ms windows and with population vectors of 10 ms bins or phase bins over a respiratory cycle.

  We observed similar results when the decoders used spike counts summed over 70 or 180 ms windows and spike counts for concatenated 10 ms or 10

  Results, The distributed representation of odor concentration is sniff invariant, paragraph 1; Figures S5G and S5H

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, The distributed representation of odor concentration is sniff invariant, paragraph 2; Figures 5E and 5F· Data missing from the dataset

  Trans-decoding of concentration falls to chance with sub-populations whose inhalation and concentration sensitivities are correlated (positively or negatively), but stays above chance with an uncorrelated sub-population.

  Thus, a trans-decoder performed at the chancelevel if it could use only the activity of a neural population with correlated sensitivity for non-olfactory and olfactory inputs (Figure 5F).

  Results, The distributed representation of odor concentration is sniff invariant, paragraph 2; Figures 5E and 5F

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, The distributed representation of odor concentration is sniff invariant, paragraph 3; Figures 5G and 5H· Data missing from the dataset

  A population of mechanosensitive neurons carries at least as much sniff-invariant concentration information as a population of sniff-insensitive neurons.

  This analysis revealed that a population of mechanosensitive neurons carries at least as much sniff-invariant information as a population of only sniff-insensitive neurons (Figure 5H).

  Results, The distributed representation of odor concentration is sniff invariant, paragraph 3; Figures 5G and 5H

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, The distributed representation of odor concentration is sniff invariant, paragraph 4; Figure 5I· Data missing from the dataset

  Over 20 consecutive 10 ms bins of an inhalation, concentration decoding within and across inhalation speeds is similar for mechanosensitive and sniff-insensitive neurons.

  Linear decoders using only mechanosensitive or sniff-insensitive neurons performed similarly throughout this inhalation window (Figure 5I).

  Results, The distributed representation of odor concentration is sniff invariant, paragraph 4; Figure 5I

  What the paper reports

  • two-way ANOVA, p > 0.05 (n = 143 neurons per sub-population)

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, The distributed representation of odor concentration is sniff invariant, paragraph 3; Figure S5I· Data missing from the dataset

  Mechanosensitive and sniff-invariant neurons have similar distributions of concentration coefficients.

  Cumulative distribution of concentration regressor coefficients for mechano-sensitive and sniff-invariant neurons (Kolmogorov-Smirnov test, P-value = 0.4793).

  Results, The distributed representation of odor concentration is sniff invariant, paragraph 3; Figure S5I

  What the paper reports

  • Kolmogorov-Smirnov test, p = 0.4793

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Airflow information affords tolerance to flow-dependent concentration fluctuations, paragraph 1; Figures 6B and 6C; STAR Methods, Simulation of sniff-induced concentration changes· Data missing from the dataset

  In a population simulated from the fitted coefficients, an inhalation-driven concentration change makes concentration decoding across inhalation speeds worse than within speed as the change grows, and adding the mechanosensory input restores decoding of the true concentration for any change.

  Consistently, decoders trained on the responses during slow and fast inhalations classified the odor concentration worse than decoders trained using only one inhalation speed as the simulated influence of the airflow on the concentration increased (Figure 6C). However, supplementing this population with mechanosensory responses was sufficient to reduce ambiguity and retrieve the actual concentration for any DIC (Figures 6B and 6C).

  Results, Airflow information affords tolerance to flow-dependent concentration fluctuations, paragraph 1; Figures 6B and 6C; STAR Methods, Simulation of sniff-induced concentration changes

  The dandiset's one file holds spike times of one mouse and none of the nasal airflow signal or inhalation times, the odor trial times with odor and concentration or the brain area of each unit, which the criterion needs; it also pools the other three concentration-series mice, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.

• ⊘BlockedResults, Sniffing causes unstable odor responses in the PCx, paragraph 2; Figures S6G and S6H· Data missing from the dataset

  Classifiers cannot decode the experimental condition from the spike waveform features of the recorded units, suggesting homogeneous sampling of cell types across conditions.

  The distributions of the waveform features were comparable across experiments, and neither crossvalidated linear nor non-linear classifiers could discriminate an experimental condition based on the spike waveforms of the units recorded (see STAR Methods).

  Results, Sniffing causes unstable odor responses in the PCx, paragraph 2; Figures S6G and S6H

  The dandiset's one file holds spike times of one mouse and none of unit waveforms, which the criterion needs; it also pools the recordings of the other experimental conditions, which are not in the dandiset. The paper says its processed data are available on request, and no public source holds them.