,
19 tweets,
3 min read
Read on Twitter
God - please, make it stop! I can't even with this shit anymore...
Okay, here goes... strap in! To begin, we have only the loosest and vaguest of hypotheses, founded on flawed reasoning (e.g., that the gut microbiome could be the source of missing heritability... like, how?)...
...or on small, out-of-date epidemiology supposedly linking maternal infection to risk of SCZ (overturned by massive studies in millions and, in any case, how would this pertain to the gut microbiome of patients...?)...
...or on vague general suggestions that the gut microbiome may in some way affect brains, though we don't know how, but there's a lot of papers on it, so SCZ is a brain thing, so maybe microbiome...
So much for the intro. Now, the design: first, assume that SCZ is one thing even though everything we know about it shows that it isn't - not clinically and not etiologically
Then analyse the gut microbiomes of a sample of people with SCZ and a sample of controls and literally look for *any differences* between them. Was there a hypothesis as to which should be different? No.
A hypothesis about the expected direction of difference? Well, no. No there wasn't. Just looking for something different somewhere. So, surely a replication sample then...
So, these differences are only correlative, but we can test causation by doing fecal transplants into germ-free mice, which is a totally physiological thing to do...
And we can test them on a whole range of behaviours, without correction for multiple tests, without a hypothesis of which should show an effect, and again without a replication sample
In this case, the ones getting a SCZ fecal transplant showed greater activity, but *less* anxiety and *less* "depressive" behaviours. Why? Who cares? You can spin these kinds of findings any way you want. We all love a good story.
Ultimately, the parallels with the transgenerational epigenetic inheritance literature are striking. 1. A supposed mechanism in search of a phenomenon... (What is the microbiome supposed to explain here?)
2. No actual mechanism.
3. No actual findings. Just lots of exploratory blips, unconstrained by prior hypotheses, uncorrected for multiple tests, and unreplicated.
4. Massive hype.
Just because the genetics of complex disorders is complex and just because the neuroscience of the highest functions of the human mind is complex, doesn't mean we need to go looking for new kinds of biology to explain them
Okay, I'm done. Sorry for the snark. But this shit gets right up my nose...
A final meta point about this kind of literature which is looking for effects rather than testing hypotheses. Do you think we would have heard of this study if the effects had been "negative"?
How many other such studies have been run that we have not heard about? How we can trust the literature if we're only seeing half the data? (The nice, shiny, apparently interesting half...)
