Day 132: Rat model validation!
Yesterday I mentioned that I had some rat data to go through. It was an old(er) dataset, about five years-old to be exact, but it was one that was going to help me validate some of my findings. Unfortunately there existed no invasive human datasets to compare my human data to, so I needed to find an animal model, in this case a rat model. Let’s discuss the importance.
Animals are not humans. This much we know is true, we’ve cured certain cancers in mice/rats only to have no effect in humans. We’ve ended countless diseases in animal models, only to have biology laugh at us when we attempt the same on humans. It has always been a perilous journey from animal models to human models. All this to say, caution needs to be taken when applying findings from rat models to findings in human models.
I know some are against animal models, but they do serve a purpose. Furthermore, for the most part, they are treated humanely. That last bit is important, most people who do research using animal models do so with reverence towards the animal and are thankful for the knowledge they help us gain. I don’t work with animal models, I work with human models, but I know people who do and they do not take the work lightly. One day we may have a better way to do things, I certainly hope we do, but until we can get to a point that we can simulate biology, animal models help us do research faster and with the least amount of harm.
That being said, if you’ve been following along with my rat model data request, or the two follow ups (here and here), you know that this was a long time in the making. Okay, not that long in the terms of research, but long enough. Recently I succeeded in getting the data from the research group I was requesting it from (THANK YOU!!!) and yesterday I took my first crack at processing the data. There were some interesting hurdles to get through, but thankfully the research group who sent me the data was very helpful in preprocessing the dataset for me.
You may be wondering what the difference was, well first this was an invasive dataset. This generally means that the sampling frequency is going to be a lot higher than the sampling frequency we use in non-invasive data collection. The highest sample rate we use in my lab is 1000 samples per second. If you think about it, that’s pretty fast! However, the dataset I was given was collected at a whopping 30,000 samples per second. Yeah, now that is fast.
Next was the other big problem. Rats don’t do what you ask them to. The data I was given was cut into 5 second chunks.This lets you synchronize the dataset more accurately with the other recording modalities, in this case EMG, because nothing is perfect so there are going to be sampling errors, this get’s compounded the longer you record. Now, I wanted ~2 minutes of rat data where the rat did absolutely nothing.
The other lab was nice enough to separate the datasets into two groups, active and rest. They also included video for each of the datasets, so I could see how the rat was behaving as the dataset was created. What I found was what they labeled active was a very active grabby rat (grabbing for food or water), but what they labeled as rest still had a rat that moved, for the most part. I managed to find a continuous dataset of ~28 seconds were the rat was stationary and not doing anything. This was the subset of data that I was after.
The lab was also polite enough to give me some preprocessing code, it was super helpful and I cannot express how grateful I was for it. Using the code I synchronized the neural data with the EMG data for each of my trials, concatenated it all together, and thus created my dataset! Now I could easily use this dataset with my own code to process it in the way that I was interested. In the dataset went and out came what I was interested in, but the big question was, would it match what I was after, would it match the non-invasive dataset that I have?
Well, I am happy to say that yes, it did! I saw the same behaviors in my resting dataset that we see in the invasive rat model. Now, that unfortunately doesn’t mean that I can share what I’m working on with you. I wish I could, I really do. However, we are one step closer to publishing my data and showing off my new technique! That is very exciting to me and when I do, you bet I will share the publication with everyone. If I have my way, it will be (of course) open access, so you won’t have to worry about paywalls. So stick around and in the next few months, you’ll see my super cool new technique that will help change a lot of things!!
Note: this assumes my technique pans out and is really real. Keep in mind that while this helps validate my claims, I still need to do more testing to make sure it is what I say it is and that will only happen with more experiments. However, my initial results are very, VERY, promising! Maybe I shouldn’t be so excited, I mean it may not pan out, but I can’t help it, I really like the research I’m doing.