I was invited to speak about "Citizen Microbiology" at the Lake Arrowhead Microbial Genomics meeting on (my birthday!) September 16th, 2014.
I decided to focus on two of the challenges associated with Citizen Microbiology: 1) the fact that microbes are invisible, and 2) the fact that the current tools for visualization of complex microbial community data are not ideal.
I present Winogradsky columns as a way to engage the public with microbial communities, particularly in a classroom setting. I also present the human face as a visualization tool for complex microbial community data.
25. 0% 20% 40% 60% 80% 100%
Water 3
Water 2
Water 1
W Sediment
G Sediment
Y Sediment
Thaumarchaeota
Gammaproteobacteria
Alphaproteobacteria
Flavobacteriia
Deltaproteobacteria
Bacteroidia
Planctomycetia
Anaerolineae
26.
27. Eyebrows Alphaproteobacteria
Left Eye Anaerolinea
Right Eye Planctomycetia
Nose Flavobacteria
Left Ear Deltaproteobacteria
Right Ear Thaumarchaeota
Mouth Bacteroidia
Jawbone Gammaproteobacteria
28. Eyebrows Alphaproteobacteria
Left Eye Anaerolinea
Right Eye Planctomycetia
Nose Flavobacteria
Left Ear Deltaproteobacteria
Right Ear Thaumarchaeota
Mouth Bacteroidia
Jawbone Gammaproteobacteria
33. …said no one, ever.
“Now, THAT’S a good looking pie chart!”
“OMG, I know I’ve seen that pie chart somewhere before.”
Your pie chart looks exactly like a pie chart I knew in undergrad.”
“I can definitely see your dad’s pie chart in your pie chart.”
“When your pie chart was younger, it looked like your mom’s pie chart but now,
it’s looking more and more like your dad’s every day.”
“Wait a minute… your pie chart looks different.
Did you have something removed from it?”
Citizen science is scientific research conducted, in whole or in part, by amateur or nonprofessional scientists
Citizen science is not a new phenomenon. The Christmas Bird count is one of the oldest, running for over 100 years.
Not only was Darwin a citizen scientist himself, but he ran his own crowdsourced research, writing thousands of letters to naturalists around the world, asking them to share their observations with him.
So, why is there such a surge of interest in citizen science?
Not only is there a surge of general interest in Citizen Science, but the contribution of citizen science to the literature is on the rise.
A small minority of these publications are projects involving microbes. Why is that? With the stupid cheap and high-throughput sequencing that’s available, why is citizen microbiology not keeping up with the pace of sequence-enabled microbial ecology?
One answer might be that the public harbors misconceptions about microbes.
I like to use the google search autocomplete feature to gauge public interest and opinion on a subject, and as you can see, the public appears to actually be receptive to microbes, this is especially striking when we compare to more a charismatic species
I have firsthand experience of the general public’s willingness to engage with the microbial world around them.
Swab surfaces for growth competition
Swab cell phones and shoes
Swab the ISS
www.spacemicrobes.org
Microbial citizen science projects do suffer from unique challenges relative to other citizen science projects
I’m going to focus on two of them, what I will term the “invisibility problem” and the “pie chart problem”. Actually, after being told I sounded too negative, I’ll call them opportunities.
First, I’ll tackle the invisibility problem. As Noah mentioned yesterday, engaging people with organisms they cannot see can be a challenging task.
But, as many of you well know, microbes are not always invisible.
Many microbes are NOT invisible. In fact, they can be quite beautiful. If you carefully pry up a hunk of aquatic sediments, this is what you might find. These beautiful, colorful layers, each of which represents a different “ecotype” of microbe. This particular chunk is from the Great Sippewisset Salt Marsh. Some of you may be familiar with this place near the Marine Biological Laboratory in Woods Hole, Mass.
These layers form along sulfur and oxygen gradients that are created my microbial metabolic processes. And these layers are easy to replicate in the lab. You just mix some aquatic sediment with a carbon and sulfur source, usually shredded newspaper and egg yolks, top it off with water, seal it and forget it
It is a long-time dream of mine to use Winogradsky columns in my research, but I haven’t been motivated enough to seek funding to do so, and not clever enough to work them into my current research.
Until now…
Our lab is currently ramping up a seagrass microbiome project, one aspect of which is to better understand the plant-microbe interactions that are taking place in the rhizosphere. Both because we want reference genomes to enable better metagenomics and because we would eventually like to have isolates with which to do experiments, we are starting to think about how to culture the microbes that may play a key role in things like sulfur oxidation or nitrogen fixation in seagrass beds.
Rather than setting up culture conditions for a wide variety of organisms, some of which are quite fastidious, we are using Winogradsky columns to do some enrichments.
Here are some examples of winogradsky columns that we have growing in the lab right now. They’ve actually survived the flight here, and I’ll pass them around now so you can take a look at them.
2, 4, and 7 weeks post-inoculation
Recipe (with diatomaceous earth) 1g each cellulose and sodium sulfate, 0.1 g each ammonium chloride, calcium carbonate, and dibasic potassium phosphate
Adding Potassium Nitrate or Ammonium Acetate changes the community composition
Western Sierra Collegiate Academy
These are so cheap and easy to create, that it is frequently done in a classroom setting.
A local AP biology teacher, Beth Dixon started building these columns in her classroom last year.
This year, we are teaming up.
Field trip to Bodega to collect sediment samples that will be used to inoculate the columns under 5 different experimental conditions that Hannah and Cassie are currently defining.
The second challenge associated with citizen microbiology is data visualization.
Microbial ecologists have approximately 3 methods of visualizing microbial community data: the pie chart, the stacked bar chart, which is just a rectangular pie chart, and the PCA plot.
I’m picking on pie charts here, but the ”pie chart problem” is a proxy for a general problem with engaging data visualization.
I will explain later why the pie chart is problematic, but first I will show you two cit sci micro projects that are doing a good job engaging citizen scientists with visual representations of their microbial communities
The first is american gut. This crowd-sourced project allows anyone the opportunity to compare the microbes in their guts to those in the guts of thousands of other people in the US and elsewhere.
Here is what David Coil received after participating in American Gut. It’s on display in our lab. It includes a bar chart, comparing the phyla in their guts to various cohorts. And Michael Pollan. Every time I look at this, I think that Craig Venter must be kicking himself for not making HIS gut microbiome the one to whom all others are compared.
It also shows some PCA plots which don’t convey much information about the taxonomic composition of your sample, but does place it in a few different contexts.
The second project is one of the Your Wild Life projects run out of Rob Dunn’s lab at NSCU. They invited citizen scientists to compare the microbial inhabitants of their belly buttons to others
Participants in the Belly Button Biodiversity project are provided with a link that takes them to a page where they can interact with their results. They present an average pie chart, and then people can choose their belly buttons from a drop down menu. Mousing over the pieces of the pie allows the taxon name to pop up. They also provide users with a series of questions that they can answer with their data, which I think is key to engagement
For Project MERCCURI, we are planning to take advantage of a web-based interactive visualization tool developed by Holly Bik back when she used to be a postdoc in the Eisen lab.
However, despite what I consider to be the best case scenarios for sharing microbial community composition data with participants in these projects, I think we can do better.
I used to think that people hated pie charts the way they hate Comic Sans. Somewhat irrational, but coming from a design/aesthetics perspective.
But, actually, pie charts are a problem because the human brain is not good at estimating area. And, it is worse at estimating angles than lengths, so the stacked bar chart is an improvement upon the pie chart for that reason. However, when the number of chunks in a bar or pie chart increase, the value of the graphic to convey information about overall community composition takes a nose dive.
So, what is the solution? I’m sure there are many, because I’m not the only one dwelling on this problem, but I think this tricky problem requires experimentation with creative, alternative solutions, and I’m going to share an example of one of those with you now.
Given that human brains are not well-equipped to interpret pie charts, how can we present microbial community data in a form of which we can make sense? The solution to this problem lies in a region of the brain called the fusiform gyrus.
This region of the brain allows us to process a particular, very complex image in an instant, undergoing relatively little decomposition into component parts. Instead, they are encoded via a holistic or integrative mechanism, as a gestalt.
The human face is the result of tweaking hundreds (thousands?) of parameters. And yet, when we look at one, we know in an instant if we’re faced with friend or foe, if that face is happy or sad. We know if we’ve seen that face before, we even know if we’ve seen that face before, but something has changed since we last saw it. We can detect similarities between faces, we can detect asymmetries in faces.
What if we could build a face that represents a microbial community? How would we do that? Well, for a first stab at it, which is what I’ll present here, we could use the relative abundance of a particular microbe to dictate a value for a facial feature.
This particular human face was designed with this software that is used for 3D animation, like for video games and movies. The facial features are controlled by these sliders, and information from slider movement is written to a nice human readable text file. Each of 100s of parameters used to build these faces slide from a value of -1 to 1
For a simple example, I will vary 8 parameters on this face. The numerical value of each parameter ranges from -1 to 1, so I have scaled the relative abundance accordingly.
Here are the 8 features.
Now, first, I’m going to show you what this face will look like if we take each of these parameters to its extreme.
For a simple example, I will vary 8 parameters on this face. The numerical value of each parameter ranges from -1 to 1, so I have scaled the relative abundance accordingly.
Here are the 8 features.
Now, first, I’m going to show you what this face will look like if we take each of these parameters to its extreme.
So, what does this look like with real data? My test case is some 16S rDNA PCR data obtained from some experimental seagrass tanks. I’m going to show you three water samples and 3 sediment samples.
So, here are the 6 samples in the form of a bar chart. What can we glean from this?
I’m just going to point out 3 things for you to focus on.
There’s more variation in the bottom three than in the top three
The sediment samples contain a greater proportion of Deltaproteobacteria, relative to the seawater samples.
The seawater samples contain Thaumarchaeota, which the sediment samples do not.
How do these data translate to the human face?
So far we’ve been looking at the representation of only 8 taxa, just for illustration.
But real data often look like this. And, I think herein lies the real value of alternative methods of visualization that take better advantage of our innate capacity to recognize patterns.
Instead of presenting people with this
We can show them this. Picking out similar faces is quick and easy, even when we have no idea what makes them similar
Or, we could hone in on a particular feature, maybe you want to group all of the guys with this haircut.
And speaking of haircuts, altering the style and color of hair, while not changing the face itself, can make a dramatic difference in appearance. We can take advantage of this by changing the color or style of the hair, or the color and style of clothing to represent our metadata, allowing people to easily group microbial communities based on their similarity to each other, and independently accounting for the metadata.
So, imagine if in the future, David’s American Gut results come back looking like this.
And, if I haven’t already convinced you that faces are worth exploring as a data visualization tool, I will close with a few quotes to get you thinking more about their potential