Connectomics for Addiction Research: Risk vs. Reward
“Connectomics” is a buzzword: knowledge on how every single nerve cell connects to each other in the brain. The word connectome refers to the exact specification of how nerve cells connect to each other in one individual organism (i.e. the fruit fly connectome, the human conncetome, etc.). This word derives from “genome,” which is the information stored in the entire genetic library.
The popular media loves this word (see most recently in The New York Times). It is an attractive unison of Big Science, smart, charismatic people, real world sci-fi, and at times gripping story arcs. Memory uploading. Immortality. Writing grants for 36 hours straight. While this makes good reading material, it is not very informative for individuals who want to actually see through what is going on, where the promise is, and where it would go.
The technology of ‘connectomics’ is essentially this: cutting up the brain in finer and finer chunks, and looking at these chunks using better and better microscopes.-Sean X. Luo
The technology of “connectomics” is essentially this: cutting up the brain in finer and finer chunks, and looking at these chunks using better and better microscopes. Researchers gather many pictures of lots and lots of chunks of brain, and computer algorithms string together the pictures into a circuit diagram of how cells in the brain are connected to each other. One might characterize interesting properties of this diagram: the way different cells are hooked up to each other might give you a hint of where or what memory is.
Researchers in basic neuroscience—the scientific discipline that studies the fundamental properties of the nervous system without consideration of solving specific applied problems—levy a particular line of criticism: the brain is more than just connections and neurons—there are the genes that control them, the proteins that live in them, and the chemicals that modulate them. The brain is too complex to be reduced to anatomy, and just examining anatomy would not be fruitful. While I think this line of criticism is valid, I think the classic reply (“you have to start somewhere”) is also valid.
…even if we have all this data, how will you analyze and extract useful information?-Sean X. Luo
Data scientists and statisticians, and increasingly administrators, have another line of criticism: even if we have all this data, how will you analyze and extract useful information? We have a decade worth of genomics data, and we still don’t really know much about how to translate genomic data into practical information that would guide medical practice. Connectomics research is expensive: billions of dollars can easily be spent on collecting a lot of data, and a plausible analysis plan should be in place first. Again I think it is a valid objection, but I am more optimistic because historically, algorithms have moved faster, and data acquisition is the main bottleneck.
Connectomics and Addiction Research: Can We, Should We?
For people like me, who live and die with patients and clinical care, the problem with connectomics is not the diagram. Or even the lack of clever algorithms.
The problem is the cutting.
I want to know what is the risk of this individual developing severe alcoholism five years down the line, when he or she is going to college and perhaps getting depressed.-Sean X. Luo
Suppose I have an adolescent who gets sent to me by his or her parents, who tell me this child is drinking a lot of alcohol and getting out of control. I want to know what is the risk of this individual developing severe alcoholism five years down the line, when he or she is going to college and perhaps getting depressed.
I am not going to cut his or her brain and get the connectome. I am not even going to get a brain biopsy.
Not now. Not a hundred years from now.
Such risks and benefits analysis, while classic in clinical medicine, is not a primary consideration for basic neuroscientists when thinking about developing and deploying technology.
Let’s consider something a bit more drastic. What if you are talking about predicting if someone is going to develop schizophrenia in five years with consistent marijuana use? Would you brain biopsy someone for that answer? Maybe, though unlikely—how does the diagnosis have the potential to change prognosis or treatment? Would you get their whole connectome? You can’t without killing this person.
This is the main difference between genomics and connectomics. Getting the genome is noninvasive. Getting the connectome destroys the connectome.
What are the alternatives? What technologies may be helpful in the future?
- One: Technology that can measure somewhat stable states of the brain and the underlying “partial connectome” that is non- or minimally invasive.
- Two: Technology that provides a way to measure the brain in real time and in natural ways (i.e. maybe “wearable” ways). How finely you cut up the brain might not matter as much as how it behaves at an ecologically relevant time-scale (i.e. days or even months and years). No current measures of the brain look at long time course, naturalistic data.
Using the Google Self-Driving Car as an analogy, you would not expect the car to drive itself or even know where it is based on a snapshot of its surroundings. It needs videos and feedbacks and prior information.
- Three: Algorithms need to be aiming at a particular kind of real world application: that the answer they spit out would effect changes in prognosis and treatment—we should take in the map and make the car drive itself. The technology should take in the connectivity within our brain and drive it toward helping people quit cocaine, if they are motivated to do so.
Where Do We Stand Today?
Currently, technology alternatives one and two are being worked on, but for a number of reasons these are not as glamorous or well funded. The third alternative is also being worked on heavily in the industry and, in my opinion, can have an enormous impact both clinically and financially, but is also largely (though not entirely) ignored by academia.
Photo Source: Pixabay