Robust and Rapid Air-Borne Odor Tracking without Casting

Near laminar airflow in behavior box. a, Heat map of the mean airflow velocity in the box as measured by anemometer. b, Heat map of the SD of airflow velocity in the box. Airflow was very stable except near the walls. c, Readout of IAA detection in the box using a PID for the five central compartments. White regions represent IAA presence. d, Learning curve for correct odor source location for all five rats. Chance accuracy level is 20% and criterion level for accuracy is 80%. e, f, Different tracking strategies based on trajectory of odor source location. e, Direct trial, whereas f represents examples of serial (left) or lateral scan (right) to find the correct odor compartment. Red and blue lines are the tracking LED positions. The red LED was present on the left side of the rat. White dots represent estimated inhalation points in the sniff cycle. Green lines project from the holding chamber toward the correct odor compartment. g, Cumulative distribution of direct and serial trials as a function of trial number, pooled across all baseline days and all rats.

Odor port versus first compartment choice distributions. a–e, Data from five rats. Scatter plots show first entry against odor compartment, for each trial. Blue dots indicate direct trials, and these are on the diagonals. Red dots represent scan trials. Data is pooled for 10 d. Vertical histograms show distribution of first compartment entry. These show a preference for a subset of compartments. Horizontal histograms below the scatter plot are distribution of odor delivery compartments. As expected, these are flat distributions. Right column, Histograms of average bin counts along diagonals, with zero bin showing diagonal line, ± 1 bins showing bins adjacent to diagonal, and so on. The central bin is much larger than the others, in all cases. The ±1 bins are larger than the other off-diagonal bins in some cases.

Baseline measures of tracking strategies of trained rats. Data shown for days when accuracy was consistently higher than 80%. a, Fraction of direct trials for each rat. b, Percentage of correct serial trials out of total number of serial trials. c, Average speeds of all rats across days for direct (D, blue) and serial (S, red) trials in the forward (F, solid lines) and reverse (R, dashed lines) direction. d, Pooled data across all rats for all days. Forward speeds (FD, FS) are significantly different from each other and from return speeds (RD, RS). No significant difference found between return speeds (*p = 0, KW Tukey HSD test. e, RMS deviation of rat trajectory from an extrapolated odor path for direct (D, blue) and serial (S, red) trials. Solid lines represent forward direction (F) and dashed lines represent return direction (R). All RMS values are statistically significantly different from each other. f, Pooled data of all five rats from all 10 d; *p = 0, KW followed by Tukey HSD test. g, Total time taken for direct (blue) and serial (red) trials. Trial time is significantly different for the two classified groups (KW Tukey HSD test, **p < 10⁻¹⁰; h). Number of direct trials: Nd, = 1851; number of serial trials: Ns, =2150. Legend 1 for a, b. All error bars for c, e, g, are in SEM. d, f, h, Box whisker plots, representing the median (gray line), 25^th percentile (bottom edge of box), 75^th percentile (top edge of box), most extreme data points (whiskers), and outliers marked individually (gray crosses).

Instantaneous deviation of trajectory from odor path for each rat. a–e, Deviation plotted for forward tracks only. Left column panels are deviations for direct trials for 1 d. Deviation values are plotted as color scale (cm). Central and right columns show examples of tracks for direct and serial trials respectively, overlaid on observed odor plumes from Figure 2c. Tracks are color-coded for each compartment: C1, blue; C2, red; C3, pink; C4, black; and C5, green. . Dashed lines represent boundaries of the box and the compartments. HC, Holding chamber; C1 to C5, compartment numbers 1 to 5.

Rats sniff actively when going forward toward the odor compartment. a, Scatter of sniff rate in forward path versus return path averaged over the entire dataset, all rats. Line is for equal rates. Forward rate is almost always faster than return. b, Sniff rate for direct and serial trials is the same. Each point is mean of rate in serial trials versus direct trials for a given rat on a given day. Line is for equal rates. c, Average sniff rate plotted against average running speeds. Blue dots are forward direction and red dots are return. These form distinct clusters. Forward sniff is faster, and run is slower, than return.

Instantaneous speed for direct trials plotted for each rat. a–e, Left column shows the speeds plotted for forward direction. Right column shows the speeds in reverse direction. The color bars show the values of the speeds (in cm/s). x-Axis is the breadth of the box (in cm) and y-axis is length of the box (in cm). Dashed lines represent boundaries of the box and the compartments. HC, Holding Chamber; C1 to C5, compartment numbers 1 to 5. The plots show multiple trials for a single day.

Instantaneous speed for serial trials plotted for each rat. a–e, Left column shows the speeds plotted for forward direction. Right column shows the speeds in reverse direction. The color bars show the values of the speeds (in cm/s). x-Axis is the breadth of the box (in cm) and y-axis is length of the box (in cm). Dashed lines represent boundaries of the box and the compartments. HC, Holding Chamber; C1 to C5, compartment numbers 1 to 5. The plots show multiple trials for a single day.

Serial and direct tracks diverge early after holding chamber. Tracks until the first entry shown for different compartments (a, e, C1; b, d, f, C2; c, C5) in direct (blue) and serial (red) trials. a–c, Non-overlapping tracks, whereas d–f show overlapping tracks for both direct and serial trials. First entries in direct trials are to the correct odor compartment, whereas first entries in the serial trials are to the incorrect compartment. Each panel shows tracks from a single session of an example rat as indicated in the plot. g, Deviation from odor plume (cm) as a function of distance from first entered compartment (cm) averaged over all rats, shown for both direct (blue) and serial (red) trials. h, Example data from two rats showing deviation from odor plume in the initial 20 cm from holding chamber. Rat B shows divergence of deviation for direct and serial trials after the 70 cm mark, where as all other rats show divergence from the beginning of the holding chamber (100 cm). i, Instantaneous speed averaged over all rats for every 1 cm toward first compartment entered. Higher speed at 110 cm indicates turning of the animals from water port. At ∼70 cm, the direct and serial speeds begin to diverge from each other.

Rats rapidly learn to track novel odors. a–d, e, g, i, The white region represents baseline (IAA) odor delivery; gray region represents Cineole, and red represents Limonene. a, Total tracking accuracy. There is a major drop on introduction of cineole, a smaller one for limonene. b, Success rate in direct trials. There is a drop only in the first few days with cineole. c, Similar plot as b for serial trials. There is a big drop at the start of cineole. d, Fraction of trials using direct tracking for all rats. This drops toward chance during the initial few days with cineole. e, Average RMS deviation across all days for all rats. There is remarkably little increase for forward direct trials. f, Whisker box plot of RMS deviation in forward direction for direct trials. Data is pooled for all rats. Nd (IAA) = 1851, Nd (Cineole) = 921, Nd (Limonene) = 724. RMS values of days for Cineole and Limonene are significantly different from IAA but not from each other (**p < 10⁻¹⁰, KW Tukey HSD test). Gray line is the median. Lower and upper edges of the box represent 25^th and 75^th percentile. Whiskers represent the extreme data points and gray crosses are the outliers. g, Mean forward and return speeds for direct (blue) and serial (red) trials for all rats in the forward (solid lines) and return (dashed lines) direction. h, Whisker bar plot of the average mean speed for direct trials combined for days with IAA, Cineole and Limonene as tracking odor. All speeds are significantly different from each other (**p < 10⁻¹⁰, KW Tukey HSD test).Note that the speed distributions are positively skewed so the medians in the whisker plots are higher than the means from g. i, Total trial time averaged for all rats for direct (blue) and serial (red) trials. j, Whisker box plot for total trial times in direct tracking for IAA, Cineole, and Limonene days. Novel odors were significantly different from IAA, but not from each other (**p < 10⁻²⁵, KW Tukey HSD test). Error bars on all line plots are SEM values.

Nostril stitch does not affect accuracy. Gray shaded area represents the day of unilateral stitch. a, Percentage of correct Direct trials out of total number of Direct trials (no effect on accuracy, p > 0.05, Student’s t test). b, Percentage of correct serial trials out of total number of serial trials. c, Fraction direct trials. d, Averaged RMS deviation pooled for all rats for direct (blue) and serial (red) trials in forward (solid lines) and reverse (dashed lines) direction. No significant differences were observed for RMS values (data not shown). e, Total trial time for all rats for direct trials (blue) and serial (red) trials. f, Total time for direct trials pooled for all rats for three groups: pre-nostril stitch (pre-stitch, N = 425), nostril stitch (stitch, N=92), and post-nostril stitch (post-stitch, N = 243). Total trial time for stitch days are significantly higher than pre and post-stitch days (**p < 10⁻¹⁰, KW Tukey HSD test). g, Mean forward and return speeds for all rats. h, Mean speeds for direct trials in forward direction. Stitch day speeds were significantly lower than pre- and post-stitch days (**p < 10⁻⁷, KW Tukey HSD test). Error bars on all line plots are SEM.

Background odor identity transiently affects accuracy of odor localization. Foreground tracking odor is Limonene (0.5%). Control (air background) is represented in white. Rats take 1-2 days to stabilize to the reduced tracking odor concentrations. Linalool background is gray shaded, Menthone is red shaded and IAA is green shaded. a, Total accuracy of all rats with background odors. There is a small drop for menthone, and a large drop for IAA. b, Fraction of direct trials for all rats. c, Accuracy of direct (d) and serial trials for all rats with different background odors. There is a particularly large dip for IAA. e, Average RMS deviation of all rats for direct and serial trials in forward and return direction. f, Averaged RMS deviation in direct trials during forward tracking. Data pooled for all rats and trials. RMS deviation with air (Nd = 524) was significantly different from days with Linalool (Nd = 295), but not from days with menthone (Nd = 241) as background (**p < 10⁻³, KW Tukey HSD test). RMS deviations for days with IAA (Nd = 540) did not differ from air (Nd = 211, KW Tukey HSD test). g, Mean forward and return speeds for direct and serial trials for different background odors. h, Averaged speeds for all direct trials during forward tracking with Linalool, Menthone and IAA as background odors. Speeds for both Linalool and Menthone were significantly lower than baseline (**p <10⁻⁸, KW Tukey HSD test). Speed for IAA (all days pooled) and air as background were not significantly different. i, Total trial time taken for different background odors in direct and serial trials. j, Average trial time for direct trials pooled across all days for air/Linalool/menthone and air/IAA as background. Air background trial time is lower than Linalool (*p = 0.03, KW Tukey HSD test). Menthone as background did not affect trial time. Average trial time pooled for days with IAA as background was not significantly different from with background air (5% significance, KW Tukey HSD test). Error bars on all panels represent SEM values.

Increased turbulence does not affect tracking accuracy. a, Image of box showing turbulent and (b) laminar flows, using smoke plumes illuminated by a laser light sheet with the source near the holding chamber. c–k, Gray shaded areas indicate days when turbulence was introduced. c, Accuracy of Direct trials and (d) serial trials. e, Fraction of direct trials out of total trials. f, Averaged RMS deviation for all rats. g, Whisker bar plot of average speed for direct trials in forward direction for days with and without turbulence (Nd laminar = 325, Nd turbulence = 522). The RMS values are significantly different from each other (**p < 10⁻³, KW Tukey HSD test). h, Forward and return speeds for direct and serial trials averaged across all rats. i, Whisker bar plot of mean speeds for days with near laminar and increased turbulence. Speed was significantly lower for turbulent days (**p < 10⁻¹⁴, KW Tukey HSD test). j, Trial time averaged across all rats for direct and serial trials k, Whisker bar plot of average trial time for direct trials during near laminar and turbulent airflow. Trial time values are significantly different from each other (**p = 10⁻³, KW Tukey HSD test). Error bars for all line plots represent the SEM values.

Trajectory changes are not triggered by odor sampling. y-Axis shows average displacement per frame since last sniff for forward track. x-Axis is time (ms) since last sniff. Each data point is a successive frame. For each plot, black and blue represents example data from different days. Star markers with dashed lines show control (see Materials and Methods), while circle markers represent post-sniff (see Materials and Methods) displacement values. Plots a and b show baseline tracking data for Rata A and D, respectively. Plot c shows pre-stitch baseline for Rat D, whereas plot d shows (Limonene + IAA background) data for Rat D. The control and sniff related displacements were not found to be significantly different from each other in all cases (Z test, Bonferroni correction, at 0.001 significance level).

Run-and-scan behavior is usually faster than casting. Color maps of time difference between casting and run-and-scan. x-Axis shows the distance to the source (meters) and y-axis shows the number of possible targets. Casting would be preferable for time difference <0 (blue shaded bins), whereas run-and-scan would be preferable for time difference >0 (red shaded bins). The black line indicates the parameters for which they are equal. a, Model 1, where rats find the correct adjacent compartment on the second try. b, Model 2, where there is no special advantage in finding odors in the adjacent compartment.

Fraction Direct performance. a, Accuracy of direct trials over 10 d of baseline training for all five rats (Legend 1). b–f, Data for all experiment subsets, ie, [baseline–novel odors (Cineole, Limonene)–novel background odor (Linalool, menthone)–pre-stitch–Stitch–post-stitch–IAA background–turbulence] from five rats are shown from b onward. Each compartment is a color-coded dashed line (Legend 2). The plots show fraction from total entries into a given compartment when that compartment was the odor port, ie, fraction = number of first entries in that compartment when odor was on/total number of first entries in that compartment