A simple GUI for analyzing BioDAQ data

The Palmiter lab often monitors food and water intake in response to a variety of stimuli. To quantify these measurements, we house mice in BioDAQ chambers, which will record how much a mouse eats or drinks down to 0.01g. While the chambers are nice, the software that comes with them is terrible. In addition to being slow, it outputs data for each cage separately, which means you get to enjoy combining Excel files. After fighting the software, I decided I could do better, and made a small GUI.

Screenshot of the BioDAQ software. Note the Windows-95 era aesthetic. You can record up to 32 scales at a time. To get a single scale, you have to unclick the other 31! Data for each cage is saved individually.

The GUI

The goal of the GUI was to be able to analyze many cages over multiple days, and output a single file containing the data for further analysis.

The GUI starts by loading a .tab file which contains the feeding data. The code for this was actually easy, as the data is just a tab-delimited text file. (It takes a little longer than I expected, a few seconds, due to datetime parsing.)

Once the data is loaded, the GUI asks the users for information about which cages to analyze, which dates, times, and how to bin the data by time. If you are interested in the data from 10 cages, binned at one hour increments, over 5 days, you can simply input those numbers. Once everything is set, you can then save the data to a .csv, which will have the same base filename as the input data. The .csv will contain columns for:

date and time
cage id
number of feeding bouts
average bout duration (in seconds)
total eaten (in grams)
number of meals
average meal duration (in seconds)
average meal size (in grams)

Left: The GUI. You can choose a range of cages, dates, times, and more. Right: Output CSV for the file given the parameters on the left. There is information for total food eaten, number of bouts of eating, number of meals, and duration for each of those. If a cage or time does not have information, the row will be blank.

If you use BioDAQs to measure feeding, and are similarly frustrated with the software, you can give this GUI a try! You only need to download two python files, BioDAQGUI.py and BioDAQmain.py. To run the GUI, you can use either python 2.7 or 3+ (the Anaconda install should have all the relevant modules). Just open a command prompt, and type:

python BioDAQmain.py

I suggest comparing the output of the GUI to some pre-analyzed data, so you can verify that it works. If you find this helpful, let me know!

A simple GUI for analyzing thermal images

One of the grad students in the lab has started a project on thermoregulation, and he measures mouse tail temperature using an infrared camera from FLIR. FLIR has an analysis tool for its cameras which works OK, but is not really designed for analyzing hundreds of images. To save him some time, I made a simple GUI for analyzing thermal images. In this post I'm going to outline the design of the gui, and how to use it, in case anyone else needs to analyze lots of thermal images.

Creating temperature images

The FLIR camera stores images in pseudocolor jpegs that look like this:

However, these jpegs do not contain the actual temperature data. To figure out where the temperature data was, I consulted this thread from two years ago about FLIR images. I learned that the data is actually contained in the EXIF for the jpeg, is only 80x60 pixels (compared to the jpeg's 320x240), contains intensity data (not temperature), and that the data was stored with the wrong -endian. Luckily, the thread contained enough details that I was able to figure out how to extract the image from the EXIF using exiftool, and was able to switch the endian using ImageMagick.

Once I had the imaging data, I then needed to convert it to temperature. Here the thread came in handy again, specifically this post which outlined how to convert radiance to temperature. All of the constants for the equation are also stored in the EXIF of the jpeg, which allowed me to calculate the temperature for each point.

The GUI

Once I was able to calculate true-temperature images, it was time to make the GUI! Previously I've made GUIs using QT Designer, but I found an outline of a tkinter script that records the x,y coordinates whenever someone clicks on an image, so I decided to modify that instead. For the image to click on, I decided to go with a grayscale version of the pseudocolor jpeg, as it looks a lot nicer.

To use the GUI, you need to install python 3.4, exiftool and ImageMagick. Then to run the gui, open a command prompt, and go to the directory with the images, then execute:

python mouse_tail_gui.py

Grayscale version you can click on! The GUI uses the .jpg for display, but loads the temperature data in the background.

Once it opens, simply click on the pixel you want the temperature of, and the GUI will output the temperature of that pixel on the command line. If you are analyzing a bunch of images, you can hit space to go to the next image, or 'q' to quit. When you are done, the GUI will save a .csv containing the names of each image, and the temperature for that image.

Exciting screenshot of a CSV! Temperatures are in Celsius.

If this GUI sounds interesting to you, you can download the script. That folder contains: 1) a README.txt explaining how to install everything, and intructions on how to run the GUI; and 2) the script for the GUI, mouse_tail_gui.py. Everything was written in python 3. If you are using the script on Windows you may have to install packages for tkinter and image. If you have any problems, please contact me, as I helped someone else in the lab set it up.

Introducing the mechanisms Twitter bot

When describing what we know about the world, scientists often have to state what we don't know. Rather than simply stating, "We don't know how X works," scientists (and especially biologists) have come up with the beautiful syntax, "The mechanisms underlying X are not yet understood." Why use five syllables when you could use 14! In a previous blog post, I explored the history of this syntax, and found its basic form dates to 1950. Later, I invented the Mechanisms rating system, the number of sentences until the mechanisms syntax is used. Since I have grown more interested in data science, I decided to look at the mechanisms syntax from a medium data perspective.

So today, I'm going to present the highlights of 25,000 abstracts that contain mechanisms syntax. Then I will introduce a twitter bot that will once a day tweet out a new mechanism that we don't understand.

Some quick stats about mechanisms

I started by querying Pubmed for 100,000 abstracts containing both a mechanism word like "mechanism" or "pathway", and a clarity word like "unknown." I then filtered the abstracts for those which used a mechanism and clarity word in the same sentence, yielding a final pool of 25,000 abstracts. If you want to see a bunch of these sentences, you can visit the Jupyter notebook for this project.

The shortest sentence in this corpus is concise, "The mechanisms are unclear."

The longest sentence highlights the beauty of the syntax, how you can write dozens of words about what we do know, then switch it up and talk about what we don't know:

Although an increasing number of studies have identified misregulated miRNAs in the neurodegenerative diseases (NDDs) Alzheimer's disease, Parkinson's disease, Huntington's disease, and amyotrophic lateral sclerosis, which suggests that alterations in the miRNA regulatory pathway could contribute to disease pathogenesis, the molecular mechanisms underlying the pathological implications of misregulated miRNA expression and the regulation of the key genes involved in NDDs remain largely unknown. PMID: 26663180

If you didn't feel like reading that, I'll summarize: micro RNAs are important for brain diseases, but we don't know how they work.

For some summary stats of the 25,000 sentences:

• 19,500 ended with some variant of "unknown"
• 8,900 included "however"
• 2,800 used "although"
• 600 used "while"

Mechanisms Twitter bot

Since I can never get enough of these sentences, I made a twitter bot to find them for me. Each day, the bot queries Pubmed for new abstracts using the syntax, and tweets it with a link.

While this was a fun side project, when you look at thousands of these sentences, they form a sort of catalogue of everything we don't know. Biofilms, leukocytes, 6-gingerol... For many of these papers, I would have to spend some time to figure out what we do and don't know.

So if you want a fun daily reminder of everything we don't know, please follow.