Wednesday, October 07, 2015

A few more words about good brains and bad brains



My previous Good Brain / Bad Brain post may have been a little out there, so here are four brief comments.

(1) HCP database.  The entire Human Connectome Project database (ConnectomeDB) is an amazing resource that's freely available (more details in Van Essen et al., 2013, 2015).


(2) Good reporting / bad reporting.  Smith et al. (2015) are to be commended for such an impressive body of work.1  But I still think it was remiss to report a population along a judgmental good/bad binary axis in a cursory manner. The correlation/causation conundrum needs more of a caveat than:
These analyses were driven by and report only correlations; inferring and interpreting the (presumably complex and diverse) causalities remains a challenging issue for the future.
...or else you're confronted with press coverage like this:
Are some brains wired for a lifestyle that includes education and high levels of satisfaction, while others are wired for anger, rule-breaking, and substance use?

“Wired” implies born that way no effects of living in poverty in a shitty neighborhood.

Oh, and my flippant observation about the wine cooler/malt liquor axis wasn't actually a major player in the canonical correlation analysis. But race and ethnicity information was indeed collected (but not used: “partly because the race measure is not quantitative, but consists of several distinct categories”).


(3) Ethics!  This brings up the larger issue of ethics. A whole host of personal participant information (e.g., genomics from everyone, including hundreds of identical twins) is included in the package. From Van Essen et al. (2013):
The released HCP data are not considered de-identified, insofar as certain combinations of HCP Restricted Data (available through a separate process) might allow identification of individuals as discussed below. It is accordingly important that all investigators who agree to Open Access Data Use Terms consult with their local IRB or Ethics Committee to determine whether the research needs to be approved or declared exempt. If needed and upon request, the HCP will provide a certificate stating that an investigator has accepted the HCP Open Access Data Use Terms. Because HCP participants come from families with twins and non-twin siblings, there is a risk that combinations of information about an individual (e.g., age by year; body weight and height; handedness) might lead to inadvertent identification, particularly by other family members, if these combinations were publicly released.

Oops.

Important Notice to Recipients and System Administrators of HCP Connectome In A Box Hard Drives

Thank you for acquiring a Connectome-in-a-Box that contains HCP image data.  This provides an easy and efficient way to transfer large HCP datasets to other labs and institutions wanting to process lots of data, especially when multiple investigators are involved. With it comes a need to insure compliance with HCP’s Data Use Terms as well as any institutional requirements.


And any participant in the study can look at the results and infer, because of their regular cannabis use and their father's history of heavy drinking, that they must have a “bad brain.” Do the investigators have an obligation to counsel them on what this might mean (and what they should do)? Yeah, stop smoking cigarettes and pot, but there's not much they can do about their father's substance abuse or their fluid intelligence.


(4) Biology.  Finally, I'm not sure what the finding means biologically. Across a population, there's a general mode of functional connectivity while participants lie in a scanner with nothing to do. That falls along an axis of “positive” and “negative” traits. And this pattern of correlated hemodynamic activity across 30 node-pair edges means....... what, exactly?

Every person's connectome is unique (“I am my connectome” for the thousandth time).2  But this mantra more commonly refers to the fine-grained structural connectome. You know, the kind that will live forever and be uploaded to a computer (see Amy Harmon's article on The Neuroscience of Immortality, which caused quite a splash).

What is the relationship between resting state functional connectivity and the implementation of thought and behavior via neural codes? This must be exceptionally unique for each person. We know this because even in lowly organisms like flies, neurons in an olfactory region called the mushroom bodies show a striking degree of individuality in neural coding across animals.3 
At the single-cell level, we show that uniquely identifiable MBONs [mushroom body output neurons, n=34] display profoundly different tuning across different animals, but that tuning of the same neuron across the two hemispheres of an individual fly was nearly identical.

In other words, a fly's unique olfactory experience shapes the response properties of a tiny set of neurons, even for animals reared under the same conditions. “In several cases, we even recorded on the same day from progeny of the same cross, raised in the same food vial” (Hige et al., 2015).

Fig. 5c (Hige et al., 2015).


I never know what to do with information like this, especially in the context of human brains, good and bad.....  Maybe: Are some fly MBONs wired for a wild lifestyle of apple cider vinegar?





Footnotes

1 Or maybe the result was a massive case of confirmation bias, as suggested in a private comment to me.

2 See this book review for an opposing view.

3 fly paper via @fly_papers (also @debivort and @neuroecology).

4 Also see Neurocriminology in prohibition-era New York.

Subscribe to Post Comments [Atom]

1 Comments:

At October 07, 2015 10:58 AM, Blogger Dwight Dickinson said...

This article is getting a lot of attention, but it hasn’t been emphasized that the amount of variance in behavior/demographics/etc explained by the combination of variables on the connectivity side is minuscule. See Figure 1c. The investigators presented many many behavioral variables and many many functional connectivity variables to an algorithm that identifies optimal dimensions of connectedness between the two datasets, and the only statistically meaningful dimension that emerged manages to account for something under 2% of the variance across the behavioral/demographic indicators. (The limited variance point may be true for the connectivity variables but I am confused about the scaling used in that part of the figure.) Maybe this isn’t surprising given the amount of data surveyed, but the specific findings here don’t seem like a big thing, as a practical matter. Certainly, there is nothing approaching an ‘explanation’ of a large positive/negative dimension of behavior — just an obscure pattern of connectivity that is associated with a tiny proportion of variance in some probably important but complexly interrelated behavioral/demographic variables.

 

Post a Comment

<< Home

eXTReMe Tracker