This is now the results from trying to predict the fly behavior doing ensembles of two predictions for the next 200 data points at two different points of the traces.

c105;;c232>TNT:

WTBxTNT:

WTBxc105;;c232:

From what we see here, there is no “flattening” in the prediction of the fly when the neurons under c105 and c232 are targeted by TNT. This is  done with around 14/15 flies for each group with two predictions in each ensemble of the two starting points. That makes a total of 15flies x 3 groups x 2 starting points for prediction x 2 predictions per ensemble = 180 prediction traces. Now I´m trying to calculate it by making correlations of bins in the prediction-observed for the same fly

These are just 5 flies (WTBxTNT) from the strokelitude where I measured the correlation coefficient on the Y-axis. In the X-axis, first bin is from 0-2 s of prediction, second is 2-4s and so on.

It seems as if some flies do nicer than others. Although it seems to me that a correlation coefficient from 0.3 isnt a big thing with all this variability. I have to find out the best binning though, I think it needs to be much more in the short term.

When I do the mean of the 5 flies measured, I do see a very slight decay. But once more I would say the decay is from the bin 1 to the second.

Here I tried another way, the RMSE, which according to literature and to my own reasoning should be a better analysis. I think RMSE measures just the differences of the absolute points whereas correlation coefficient is rather if the direction and degree of variation correlates (covariates). I find a very weird result. The fit is bad, the it gets better (but it should be just a chance event because correl coef decreases) and then it get very bad and so on.

I think for the future I have to make ensembles of two k neighbours maybe, which seem to increase the prediction power 10-15%. And maybe not look that much into the future as it was done here (10s).

Here some examples of predictions vs observations:

For a further data analysis, we should have a minimum number of samples. During the last week, I have been measuring the wingstroke amplitude of the flies, to get at least 10 samples of each one(two controls and the experimental line). Here three examples of the different lines:

Males WTB x C105;;C232 (control)

Males UAS-TNT-E x C105;;C232

Males UAS-TNT-E x WTB (control)

After getting good results in the measurement of the wingstroke, and solving problems with the sampling intervals (image below). I have started to measure flies for the experiments.

The time between samples was different depending on the background programs running on the background:

Figure 1

More differences in time within samples when more programs are running at the same time as strokelitude (3nd 1/3 of the plot), when just the display of the camera is running( 1st 1/3 of the plot), and everything shut down(2nd 1/3 of the plot).

The differences among sampling intervals was bigger but with an adjustment of the data, Christian Rohrsen managed to changed. The time between two samples could arise until 1.7 seconds and with the correlation, the time is not bigger than 0.05 seconds.

The strains of flies used fore the experiments are:

UAS-TNT-E, that express the tetanus toxin in the neural cells, using GAL4 system. Being Used as a control.

c105;;c232, that contains a promotor region to express the toxin. Being used as a control.

And the cross between both to have the expression of the tetanus toxin.

Some data from the control flies (UAS-TNT-E):

And some data of the spikes:

This is just a proof of concept how useful the interpolation is for this purpose: and I would say it´s almost useless. If I delete from the raw data (1st graph) some pieces (2nd graph), and I make a spline interpolation (3rd graph) it doesn´t match that well. Linear interpolation was done in previous weeks with not much of a success. Anyway, the result of a linear interpolation can be imagined by eye just by joinning the two ends with a straight line, and this doesn´t occur in the fly behavior as we can see in the raw data (1st graph).

There could be two posibilities for spike detection: one is the one from Ute and the other one is the one from Maye. It seems to me that the one from Maye is more precise. I have run the script but I do not get so many spike detections as he gets. I did try several thresholds for spike detection and doesn´t change very much. So I have to work more on it to see what is really the important factor for a proper spike detection

After the measurements of this week, I get better results in the trace of the fly;

(1)

With the corresponding trace exerpt with the spikes of the figure above (1)

(1)

and two more from other flies:

After the results in tracing of the flies from last wee. This week I’ve improved detection of the wings and the care of the flies. The results seem similar to the flight of a fly. Attached we find the trace Downsampled of the best four flies I got:

Those are the results of measuring the wingstroke amplitude with Strokelitude and “The ping pong ball machine”. And analysed with R studio and the scripts written by an student.

The results are not too good, comparing them with the result obtained in other experiments.

Two pictures of the trace down-sampled of my flies:

and in the next picture is how it should be plotted: