Slowly getting the sample size going. As of now, it seems aPKC is either not needed in steering motor neurons b1 and b3, or that knocking aPKC out in only one of them is not sufficient to have an effect on operant self-learning. Shown is the first 2min test period after 8min of training, all three groups seem to show learning, at least at this stage:

It looks like panneuronal downregulation of the foraging gene using the elav-Gal4 driver line impairs self-learning in the flight simulator.

Now that we have established that the plasticity underlying self-learning is located somewhere in the steering motor neurons of the ventral nerve cord, the next question is: which of the neurons are involved. To this end I have now started to knock-out aPKC in either B1 neurons or in B3 neurons. The muscles innervated by these motor neurons are an agonist/antagonist pair and serve to advance/delay the turning point of the wing, leading to a larger or smaller, respectively, wing stroke amplitude. Asymmetry in the activity of these neurons leads to yaw torque – which is the behavior we condition. In the first two weeks, I noticed that all three groups (B1- knock-out, B3 knock-out and genetic controls) seem to fly reasonably well. So far, it doesn’t seem like there are any striking differences between the lines, but it is still early days and about three times more animals are needed before one can draw any firm conclusions:

Slowly the data are filling up and we start to see some differences emerge between the controls and the aPKC knock-outs:

We still need to get to about N=40, so there is still some way to go.

Going over the optomotor responses with a fine comb revealed a bunch of flies where the algorithm wasn’t able to provide a proper fit for the OMR asymptote. Therefore, I will need more time to finish the data set. Here the current torque-learning PIs:

Clearly, the genetic controls learn while the flies with knocked-out aPKC in FoxP neurons fail to show a significant learning score. However, the OMR asymmetry effect in the genetic controls appears weaker than the one we discovered in WTB flies, as can be seen in the OMR traces after the self-learning:

Then again, at the .05 level, the asymmetry index is significant. Not the alpha level we commonly use, but also a lower N than we strive for (above is before training, below is after):

The transgenic experimental flies, in contrast, don’t seem to show much of an effect at all:

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

Just to have an example of yaw torque datasets and how they should avoid:

Before collecting the actual data I make sure that the flight simulator in the basement does its thing ;) Here are the test runs with wtb flies for different protocols:

—>> yaw torque

–>> switch mode

–>> habit formation

Finally have about half the number of flies needed. It looked like the flies that used the FoxP virgins didn’t fly as well as the other flies, so we dropped that branch and have stopped using them for the crosses. Pooling the FoxP>aPKC/CRISPR flies no increases the N in this group:

Added more flies to the aPKC knock-out in FoxP neurons. Now the knock-outs are close to zero, but one of the controls, too. Still too early to say much. The figure looks ok now, but the Bayes Factors get chopped off. need to fix this. As I’m, already working on some figures (histograms) in the code, I can fix this as well.