Last week, Pavan Kaushik and I finally got the VR-panel setup working. I placed four flies in the machine, one in each VR cube, and then ran each fly three times with a sequence of 8 different pattern wavelengths, also repeated three times, each one rotated both counter-clockwise and clockwise. This is also the sequence which we will use in the course. This is what the data should look like:

and this is what our data looked like:

Looks to me like we got it working on the first try. Love that data. Great example to compare the course data against.

The code for evaluating the VR-Panel data is on GitHub.

Because the effect of yaw torque training on optomotor responses (OMRs) is still very small for now (we work on improving that), I pooled the two groups in which aPKC was knocked out in either motor neuron (MN) b1 or MN b3, as both these two groups and their WTB x aPKC/Cas9 controls seem to learn just fine (torque preference text after 8 minutes of training):

Obviously, we still need to check the Gal4 driver lines are really targeting the right neurons, but assuming they are ok, it seems like neither an aPKC knock-out in b1 alone nor in b3 alone is sufficient to affect operant self-learning. Maybe this is due to b1 and b3 acting as an agonist/antagonist pair and if one of them fails to show plasticity, the other is sufficient on its own? Another explanation could be that the torque preference depicted above is mediated by other neurons than b1 or b3 and that the OMR modulation is gone in these flies. Because the OMR effect is so small, I pooled the two groups, threw out all flies that didn’t have at least an acceptable OMR and halfway accurate OMR parameter estimation and plotted the OMR traces of the remaining 35 flies after training:

So despite these flies learning well, the OMR does not seem modulated as one can see in WT flies. However, there my be a slight effect for the fly punished on right turning torque, perhaps? However, this group also has much larger errors, which I would need to check the reason for. The quantification of the OM symmetry does not show any hint of an effect, though:

Below the total evaluation before and after training. What is weird is that despite there being no effect after training, the correlation between torque preference and OMR asymmetry seems to be there – or is it just the three outliers?

Either way, when I pooled the control flies from this experiment with the same genotype from the last experiment to get to 42 flies, only the group that was punished on left-turning torque showed the modulation:

Accordingly, the quantification shows no difference ion the control group either:

And no significant correlation between the indices either:

All in all rather puzzling results that reinforce my view that the OMR effect is much too small to practically work with. That means one of the next goals must be to get this effect size increased by, perhaps, decreasing the strength of the optomotor stimulus?

Now with over 20 flies in each group, it becomes more and more apparent that both the flies without aPKC in either b1 or b3 steering motor neurons still learn just fine:

As with the aPKC knock-out in FoxP neurons, also here, the optomotor response seems normal as well:

Interesting is the scatter in the slope parameter for the control flies:

Not many fliers left now. Will start evaluating optomotor asymmetry now.

There had been some concerns about the optomotor display in the group evaluation sheets showing right-turning torque on the left side of the graph and vice versa. Also, the use of standard deviations seemed to blur differences between the experimental groups:

Because of these concerns, I have swapped the traces and used standard error of the means instead of standard deviations:

What do you think? Better or worse? Feedback very welcome!

I now switched the sign of the Optomotor Asymmetry Index in flies that were punished on producing right-turning torque, such that weaker punished torque shows up as a positive index. After that was done, I plotted the correlation between the optomotor index and the preference index:

I had to get rid of eight flies where the optomotor response was already asymmetric before the training started, so now I only have 33 flies. But with these flies, there is no correlation before training and a very significant correlation after training.

Any suggestions about appearance of the graphs?

Would it be useful to plot optomotor and performance indices as raincloudplots next to the regressions?

This would be the complete figure:

The results of comparing optomotor responses after self-learning remain solid. There still is a small asymmetry between the left/right groups, but nothing dramatic:

The offspring of my first experimental fly cohort finally hatched! Below you find a few first pre-tests I ran last week :)

First, here are the results of a quick test to see if the offspring shows a preference for the parentally trained side after the first training period:

After that I played around with the laser a little bit to find the learning threshhold. I set the laser on 2,6V but the results I got look a bit weird:

-> untrained wtb flies

-> offspring of trained wtb flies

I´m optimistic however there is still a lot of work to do…