I think it’s important to recognize not only how fake science can degrade medicine and exploit health care consumers, but also how real science can benefit medicine and consumers. It’s also important to separate hype from reality, because there often is science-based snake oil, meaning that there are fake treatments based on the hype of real science.
We are near the beginning of some real and exciting medical science and it’s important to track where we are. We are just beginning to see the benefits, for example, of CSRISPR technology. Genetic engineering in general is poised to transform many areas of medicine. Another important emerging area is our ability to map, interface with, and alter brain function. Neuromodulation is a legitimate area of treatment, but also ripe for exploitation.
An exciting area of research is the large collaborations trying to map the human brain in detail – the connectome. This is often likened to the genome project, but it is a much greater undertaking. The ultimate goal is to create a complete wiring diagram of the entire human brain. Of course, such a diagram, if represented in computer hardware or even virtually in software, can be functional. A complete wiring diagram could be a functional digital brain.
In fact, it is difficult to have the first without the second (at least in pieces, if not all together). In order to know if a wiring diagram is accurate we need to know that it functions like a brain. A recent series of publications illustrates the process nicely.
This is the work of The Machine Intelligence from Cortical Networks (MICrONS) Project, a collaboration t of more than 150 scientists and researchers from the Allen Institute, Princeton, Harvard, Baylor College of Medicine, Stanford and many others. Their goal is to create a “dense reconstruction of the structural connections and functions” of the human brain.
Recently they have accomplished this for a small piece of the mouse visual cortex. They have created the largest functional wiring map of a mammalian brain to date “containing more than 200,000 cells, four kilometers of axons (the branches that reach out to other cells) and 523 million synapses (the connection points between cells).”
The process they used was to take a cubic millimeter of mouse brain and sliced it into 25,000 layers, each 1/400th the thickness of a human hair (which always seems to the be standard metric by which such things are communicated to the public). Another team then used artificial intelligence (AI) to reconstruct these layers into a 3D model.
As an aside, AI is yet another area where it is important to separate hype from reality, but I see both kinds of error being made. It is easy to overhype what AI can currently accomplish, but just as often I see dismissing the power of current AI applications as hype. Meanwhile, AI is transforming research in many areas, and neuroscience is one. AI is great at making sense of massive sets of data, so it is ideally suited for areas of research such as genetics and neuroscience.
The result of this collaborative effort was a details map of a piece of the visual cortex of the mouse brain. But they needed that map also to be functional, to see not only how the cells were connected together but how they functioned. So they trained AI on information from mice whose brain activity was being recorded while they were watching action movies.
This enabled the scientists to see how the wiring worked, which lead to new discoveries, such as the role of certain type of inhibition in brain function. They found that inhibitory cells were not just downgrading overall neuronal activity, but some were highly selective in which neurons they inhibited, meaning they were a functional part of the information processing (not just brakes).
To test their new model they used it to predict how the mouse visual cortex would respond to specific stimuli, and it was able to do so. Crucially, it was able to predict the response to novel visual stimuli, meaning different than the training data.
In essence we now have a working digital model of a piece of the mouse visual cortex. This is an important proof of concept. Their plan is to keep going, and to progress from sensory cortex to higher level processing, abstraction, planning, and other functions. The ultimate goal, of course, is to create a fully function digital mouse brain. And once that is done, to complete the same project for a human brain.
But even with just a mouse brain, that can serve as a model for a generic mammalian brain, and allow us to study virtually many types of neurological disease. We can compare healthy brains to diseased brains and see what’s different. How is the wiring pattern of an autistic brain or schizophrenic brain different than a typical brain?
Imagine having fully functional digital mouse brains – this could enable us to run psychological and neurological experiments “in silico” orders of magnitude faster and cheaper than having to deal with live animals.
Of course, replicating this project for a human brain introduces very tricky ethical considerations. We need to think carefully and work out these ethical implications now, before the technology is upon us, which will likely be sooner than many people think.
