Pioreactor dev blog #17 - Noisy data

Hi! My name is Kelly, and I’m the new hire here at Pioreactor! I’m a recent graduate from the University of Waterloo with a BSc, and I’ll be working on the experiment repository, while also making some blog posts to document the experience.

Over the last two weeks that I’ve been here, I’ve been thinking about a whole variety of tests we can do with the Pioreactor (and adding them to our repository). We’ve started with some basic experiments, changing parameters such as temperature and salt content to observe any changes in growth rate in yeast. 

This is not without its hiccups. We noticed some peculiar noise that occurred for only a few minutes at a time in some yeast vials. This is a new occurrence that didn’t happen at the previous office location (yes, we moved!).

Identifying the issue

One of the more challenging hurdles to overcome in R&D is the identification of sources of error; Is it a hardware issue? Is there something wrong with the electricity in our building? Could it be biological? 

To narrow it down, we’ve noted that this noise seems to build up over time, and persist mostly at the end of the log phase of growth… Could this be a biological variance?

Figure 1: An example of our noisy data.

An idea that Cameron had: perhaps there was potential build up of gas in our vials. Something to know about yeast is that they’re fermentors. The ability to ferment glucose into ethanol and carbon dioxide makes it a key component in the creation of food items like bread and beer. The key accumulation here is carbon dioxide  — CO₂!

Note that bubbles arise in an anaerobic environment, since without oxygen, the concentrations of CO₂ build up quickly. We can see this when comparing the products for aerobic respiration versus anaerobic fermentation: 

Aerobic: 6 O₂ + 2 C₃H₄O₃ → 6 CO₂ + 6 H₂O + 38 ATP

Anaerobic: 2 C₃H₄O₃ → 2 C₂H₅OH + 2 CO₂ + 2 ATP 

In an anaerobic system, 2 CO₂ molecules are produced for every 2 ATP molecules (used by yeast for energy). This is basically a 1:1 ratio! In contrast, aerobic respiration creates 0.15 molecules of CO₂ of for every 1 molecule of ATP. Anaerobic systems produce over 6 times more CO₂ for the same amount of ATP. 

We ran a basic experiment again during the day to try and catch what we believed were CO₂ bubbles introducing noise in our OD readings, and lo and behold:

Figure 2: Bubbles were detected!

Potential solutions

Since we’ve identified (with reasonable confidence) that there was an accumulation of CO₂, we modeled the stirring with some bubly lime soda.  

Figure 3: I didn't take a picture, but it basically looked like this with a stirbar instead of a finger.

We noticed that bubbles tend to build up on the stir bar (nucleation) and funnel up through the vortex produced by fast stirring, thus leading to the accumulation of noise at the end of the log phase. With closed caps, oxygen is deprived from the system, resulting in yeast anaerobic fermentation and extra production of CO₂ (gas bubbles). Interesting to note is when the caps are extra tight, no CO₂ bubbles are produced since the pressure allows it to remain dissolved in the media.

One of the ways we can counter this is by using a baffle. We tested this by inserting a stick to introduce more randomness in our mixing. We also tried slowing down the stirring. Both methods worked  — but we settled on the easiest prevention of bubble formation: changing caps (to those with tubing) to allow some air circulation to occur. The yeast will now aerobically respire. There are some caveats in that the system is no longer as sterile as it could be, but we haven’t noticed any contamination thus far. 

We will continue doing experiments on yeast (and other microbes! More to come), and building the repository as the experiments are performed.