Week 3: Preliminary OKR Analysis and More Data Collection

Mar 05, 2019

This past week, I’ve been analyzing all my data from those 17 OKR experiments.

During the course of an experiment, mice will make a lot of eye movements that are extraneous or unrelated to the stimulus. Normally their eye movements are sinusoidal, following the stimulus in the visual field (as part of the optokinetic reflex, the OKR), but sometimes they will shift their field of view to a different point, or blink, which results in drastic eye movements, which we call saccades.

To combat this, we pick out the sections of the good data from the bad in a process called desaccading. Our resident coding expert Max has helped write an amazing magical program that looks at all our data, identifies bad sections with saccades, and throws them out. Below is a sample trace of the filtered OKR you saw last week. It shows us what it filtered out by turning the saccades black and keeping good data in blue. It also fits a sine curve to our good data, and this sine curve can be analyzed for an increase in eye amplitude over time. If the amplitude of the sine curve increases over time, that shows us that the mouse is moving its eye more.

We set thresholds for what we want to call a saccade and what we don’t. It’s important to keep thresholds the same across all the data — that needs to be a controlled variable. We spend quite a bit of time playing around with it to make sure that the threshold we set is the best one for the greatest number of data points.\

Aside from this OKR data analysis, I’ve been running some more experiments on these mice (you’ll hear about them next week). These experiments take an average of 1.75 hours, so it’s about 30 hours of data collection each week. Another member of the lab also has to use the experiment rig, so we have created a calendar blocking out who gets which time. I’ve learned how important it is to communicate with others when sharing equipment; it eliminates nasty surprises and gives me a better sense about how much I’ll be able to get done in a week. We also communicate about whether there is anything wrong with the rig (and that happens pretty often, especially with the wires that communicate information about the mouse’s eye movements). Fixing it requires serious precision soldering skills (it’s recommended we solder this thing under a microscope), and I like to think I’ve developed mine well, although my track record probably suggests otherwise. 🙂

See you all next week! Thanks for reading!


2 Replies to “Week 3: Preliminary OKR Analysis and More Data Collection”

  1. Beryl Z. says:

    Super interesting! I’m curious as to what kind of stimuli you use to make sure that the mice are focused– do you ever switch up the types of stimuli?

    1. Jaydev B. says:

      Generally, we keep them alert with a light flash between each time block of training. But they are also surrounded by the drum (see last week’s video), so that’s all they can see. They can’t really get very distracted.

      Any switching up of stimuli that we do would have to be randomized, so we tend to try to avoid that.

Leave a Reply

Your email address will not be published. Required fields are marked *