Big Data looks at your poop
March 8, 2015 9:57 PM   Subscribe

The team gathered samples over the course of a year from sewage treatment plants in 71 different cities in 31 states, chosen for their geographic spread and range of obesity rates. The leanest city sampled was Steamboat Springs, Colorado, with an obesity rate of 13.5 percent, while the heaviest was St. Joseph, Missouri, with a rate of 37.4 percent. ... "If we had a city with a higher percentage of obese people, we would see a higher percentage of the kinds of bacteria associated with obesity,"
posted by mecran01 (31 comments total) 13 users marked this as a favorite
 
Oh wow, this is fascinating!
posted by Blue Jello Elf at 10:21 PM on March 8, 2015


MetaFilter: Oh, wow, this is fascinating shit!
posted by five fresh fish at 10:31 PM on March 8, 2015 [16 favorites]


Not a scientist, but I have a skeptical reaction. "Obesity rates" as a category is problematic. And categories of bacteria, that seems pretty broad.
posted by ClaudiaCenter at 11:20 PM on March 8, 2015 [1 favorite]


Mark my words, in 20 years Japanese toilets will automatically monitor your gut microbiome and post the results to social media.
posted by kscottz at 11:31 PM on March 8, 2015 [8 favorites]


I neglected to add that you can donate your poop to a large scale study.
posted by mecran01 at 12:32 AM on March 9, 2015 [2 favorites]


Drilling down to the actual study (which is not only full text but Creative-Commons-licensed) it looks to me as though they took a couple hundred samples of sewage and then used gene sequencing equipment and analysis software to gather and sort the resulting individual gene sequences for species identification, grouping of mutations and other genetic variation within species, and statistical analysis of the distribution of all those things. And the software they used appears to be open source! Cool stuff.

And now I'm immersed in spelunking through articles on the basics of DNA sequencing technology on Wikipedia. I wonder if my little toddler cousin will be doing smaller-scale versions of stuff like this for high school science projects.
posted by XMLicious at 12:34 AM on March 9, 2015 [2 favorites]


Mark my words, in 20 years Japanese toilets will automatically monitor your gut microbiome and post the results to social media.

Already been done.
posted by mikepaco at 2:02 AM on March 9, 2015


"Obesity rates" as a category is problematic.

The researchers used first and fourth quartiles to assign members of studied populations to "most lean" and "most obese" groups. Using standard deviations away from the mean yielded similar results.

Median statistics tend to be robust to outliers, and applying a different method to see if the same answer comes back seems reasonable.

And categories of bacteria, that seems pretty broad.

This is a field of research, and the researchers in this case looked at the `“core” gut microbiota among human populations in the United States`.
posted by a lungful of dragon at 2:03 AM on March 9, 2015 [8 favorites]


Now I'm waiting for cities to brag about their fecal type.
posted by pracowity at 3:03 AM on March 9, 2015


ClaudiaCenter: "Not a scientist, but I have a skeptical reaction."

This is almost never a good way to start something.
posted by barnacles at 3:25 AM on March 9, 2015 [13 favorites]


IMO that's actually a pretty good way to start reacting to a piece of science journalism, which half the time is a sensationalizing rewrite of a press release written by a university or company's marketing department.
posted by XMLicious at 3:47 AM on March 9, 2015 [9 favorites]


Now I'm waiting for cities to brag about their fecal type.

Next up on Fecal Deathmatch: Bayonne vs. Staten Island
posted by mikelieman at 4:43 AM on March 9, 2015


Is this a real scientific study, or were they just going through the motions?
posted by Devonian at 5:35 AM on March 9, 2015 [9 favorites]


the more i think about this, the less i understand why they did this, except as part of some larger social mania for surveillance.
posted by ennui.bz at 5:58 AM on March 9, 2015 [1 favorite]


Now every time I take a dump I am going to yell "Big Data!"
posted by srboisvert at 6:01 AM on March 9, 2015 [9 favorites]


the more i think about this, the less i understand why they did this, except as part of some larger social mania for surveillance.

Or perhaps the possibility that, given sufficient data from different environments and sufficient computing power, it may be possible to determine connections between gut microbiome and various medical conditions, to get a better understanding of the gut microbiome and its interaction with the organism, ultimately leading to greatly improved diagnoses and treatments for various conditions affected by it (of which, it is turning out, there are increasingly many).

In contrast, phrasing this in terms of the dystopian bugaboo of surveillance makes little sense; aggregate data for frequency of various types of bacteria in a city's sewer system is not very useful for identifying the lifestyles or health conditions of individuals.
posted by acb at 6:24 AM on March 9, 2015 [16 favorites]


Now every time I take a dump I am going to yell "Big Data!"

And every time I hear someone say "data dump", it's going to be even harder not to giggle!
posted by Blue Jello Elf at 7:06 AM on March 9, 2015 [8 favorites]


aggregate data for frequency of various types of bacteria in a city's sewer system is not very useful for identifying the lifestyles or health conditions of individuals.

This generalizes to a very important baby/bathwater-type comment, since the dystopian bugaboo of surveillance is real and so is the utility of digging through vast piles of anonymous shit. In fact, it would be great if the rest of the data we constantly emit became as anonymous upon aggregation as our more corporeal emissions.

the more i think about this, the less i understand why they did this, except as part of some larger social mania for surveillance.

Surveillance would be a team harvesting shit in the sewer and working out when it got there, while another team keeps track of which washroom lights were on in the city, and attempts to correlate illumination with new arrivals to the sewer to deduce who produced what and when. Traffic analysis, kind of...
posted by busted_crayons at 7:08 AM on March 9, 2015 [3 favorites]


It's not just for obesity! Wastewater analysis can be used to see the general trends of drug use in an area.
posted by ymgve at 7:40 AM on March 9, 2015 [1 favorite]


If "they" want to surveil your poop, it makes more sense to put detectors in your toilet, or just under it in the pipes. Why "they" would want to do such a thing, I have no clue, but combing through an entire city's worth of sewage has got to be the least-effective way of doing so.

It's a pretty good way to learn about your city's overall health, though.

I don't find the obesity thing offensive, because there does seem to be some promising research on gut flora and weight gain/loss out there, and you can pursue that while still also resisting the discrimination and ugliness that gets directed at people perceived as overweight.

Not to mention the c.diff sufferers and Crohn's sufferers and others who could really use some better solutions for their gut dysfunctions.
posted by emjaybee at 8:11 AM on March 9, 2015


Surveillance would be a team harvesting shit in the sewer and working out when it got there, while another team keeps track of which washroom lights were on in the city, and attempts to correlate illumination with new arrivals to the sewer to deduce who produced what and when.

It'd actually be pretty easy to isolate a single building's sewage. We do a lot of sewer rehab at my office, and use robotic equipment to inspect and repair individual connections. Some minor modifications to the equipment would allow someone to collect the entire waste stream into a truck sitting at the next downstream manhole.
posted by Ham Snadwich at 8:27 AM on March 9, 2015


obligatory
posted by toodleydoodley at 8:38 AM on March 9, 2015


Here's a related funn fakt about weight and shit microbes:

A team of scientists took some shit from skinny human ladies and scientifically shoved it up the buttholes of little lab mice. These mice reportedly maintained their svelte Minnie Mouse figures. Using additional science, the scientists shoved shit from fat human ladies all up into different mice. These mice got fat.

Therefore skinny people shit is the miracle obesity cure of the future. You just put it up in there and the pounds will melt away because bacteria.

Alternatively if you're on the road, and you stop to take a shit at a McDonald's or an Arby's and the toilet water splashes up into your butt--as it is wont to do--it is probably going to make you fat like most of the people that take shits there. Science.
posted by dgaicun at 8:42 AM on March 9, 2015 [2 favorites]


Therefore skinny people shit is the miracle obesity cure of the future.

Oh, but there is a complicating factor! Another team of scientists took two strains of mice -- one which is confident, calm, and generally suave, and the other of which is as anxious and as nervous as a closeted congressman in an airport bathroom. They did fecal transplants between them, and the suave mice got anxious and the anxious mice got calm. Now because this is real science, the researchers did not conclude that POOP CURES ANXIETY, but rather took the much more conservative view that perhaps donors for fecal transplants should be evaluated for mental as well as physical health.
posted by KathrynT at 9:30 AM on March 9, 2015 [3 favorites]


I believe I read recently that a lady receiving a fecal transplant was suing because she gained a significant amount of weight since the transplant, but I'm not trying to google that at work.
posted by Ham Snadwich at 9:34 AM on March 9, 2015


You were probably remembering this story, which hit the news last month.
posted by Blue Jello Elf at 10:45 AM on March 9, 2015


It'd actually be pretty easy to isolate a single building's sewage. We do a lot of sewer rehab at my office, and use robotic equipment to inspect and repair individual connections. Some minor modifications to the equipment would allow someone to collect the entire waste stream into a truck sitting at the next downstream manhole.

It looks like my neighbours and I are going to need to set up some kind of Blinfolded Bucket Brigade Tor Project.
posted by busted_crayons at 11:13 AM on March 9, 2015 [1 favorite]


Sorry, hit "post comment" by accident so this is a DO OVERRRR

Not a scientist, but I have a skeptical reaction. "Obesity rates" as a category is problematic. And categories of bacteria, that seems pretty broad.

I'm a scientist and I work on gut bacteria so hopefully I can respond usefully here. Others have addressed that they binned obesity into quartiles, which would be robust to outliers. The authors also have a wider geographic sample than many previous gut microbiome studies (the first round of the Human Microbiome Project only sampled from two areas in the USA, if I'm remembering right), which should help to control for other sources of variation.

It's true, though, that we don't know whether any associations they find are causal. There are other confounding variables that could be correlated with obesity, with probably the foremost among them being diet. Since diet is known to rapidly alter the microbiome (link), and variation in diet wasn't measured here (and is arguably harder to measure quantitatively than BMI), it could be that some of the associations they find are linked indirectly to obesity, rather than directly. Nevertheless, their claim is not so much that these bacteria contribute to obesity but rather that they are biomarkers associated with obesity, which I think is not at all overstating what they found: you don't need a causal relationship to do a good job of prediction.

Actually, the association they find here is interesting because it's kind of the reverse of what you might expect from some earlier literature on the relationship of gut flora to obesity, but to explain that I need to explain the "categories of bacteria" statement first.

The categories they're looking at are taxonomic categories, so groups of organisms that are related by descent from a common ancestor. There's actually huge swings between individual gut microbiomes at the phylum level (and for comparison, humans, insects, and tunicates are all members of the same phylum). Most gut bacteria fall into one of two phyla, Bacteroides and Firmicutes, and it was thought for a while that the ratio between these two phyla was associated with obesity. It was a tempting hypothesis because this was observed both in experimental mouse studies as well as some observational human studies, but it turned out that the observations in humans haven't really generalized across studies (link). This probably means that taxonomic classification alone is not quite adequate for understanding the relationship between obesity and the microbiome (which remember we have experimental, not just observational, support for).

Anyway, the researchers here were actually able to go much farther down the tree of life: their taxonomic data uses 16S sequencing. 16S is a super-conserved gene that encodes a subunit of the ribosome, which is part of basic cellular machinery, and is therefore something that basically all bacteria share. However, the gene does have lineage-specific mutations, and so by sequencing it you can differentiate between bacteria at around the "species" level. The concept of "species" in bacteria is kind of slippery to begin with, so people usually use "operational taxonomic units" or OTUs instead, which are often defined in practice based on some cutoff of 16S similarity. Here, the researchers actually use a new, fancier method based on Shannon entropy, which is a concept from information theory that should allow even finer-grained resolution at the taxonomic level. So they're actually measuring taxonomy pretty precisely here.

That said, even species-level taxonomy doesn't capture everything important about a microbe, and bacteria in the same species can have pretty different gene content (link), meaning that there may be a lot of genes that are not shared even between members of the same bacterial OTU. To get around this, you can look at genes instead of species/OTUs/oligotypes as the operational unit (so you'd be saying "this bacterial gene is associated with XYZ" instead of "this species etc.") using a type of technology called shotgun sequencing. However, it's way more expensive to do shotgun sequencing: if you think about it, doing taxonomic profiling with 16S requires only sequencing a single gene per genome, and now you need to get thousands. So it's not surprising they didn't do it on a dataset this big, but it could have been very informative if they had.

Back to the obesity association: what they find here is actually that several OTUs in the general "Bacteroides" phylum are actually more abundant when the obesity rate is higher, with some OTUs in the "Firmicutes" phylum having the reverse trend. This is kind of the opposite of what was noted in some studies, where Firmicutes were associated with greater obesity risk than Bacteroides. Having said that, because the history of taxonomic associations with obesity from observational data is kind of fraught, I don't think that's a reason to particularly either doubt or believe that these associations would hold up in an independent sample.
posted by en forme de poire at 12:39 PM on March 9, 2015 [11 favorites]


Great comment, en forme de poire. Thanks.
posted by a lungful of dragon at 3:32 PM on March 9, 2015 [1 favorite]


Why can't "big data" just mind its own business?

Oh, I remember: It beats working on something that will better things for all of humanity.
posted by InsertNiftyNameHere at 11:49 PM on March 9, 2015


This really is merely "big data" as in "very large quantities of data", in that the gene sequencing machines used to analyze all of the different microorganisms in a couple hundred sewage samples produced untold reams of information which required intensive computing power to analyze. So not "big data" in the sense of mining consumer transactions and web/app clickstreams and other records to profile and predict human behavior, but something that applies to much more conventional scientific and medical endeavors.
posted by XMLicious at 12:20 AM on March 10, 2015 [2 favorites]


« Older 'You're on your own.'   |   The hopalong attractor Newer »


This thread has been archived and is closed to new comments