Neuro - Big Trends

Catching up on comments for recently added papers. Back on 4/26 added a *bombshell* paper from Ioannidis and collegues on power and research quality (Button et al., 2013, in NRN). First, the paper summarizes some statistical points which *should* be well known by now: small samples sizes lead to poor power and extreme levels of sampling error in estimating the real effects of a treatment. With small sample sizes, the occasional over-estimate of effect size will lead to publication, but this will be misleading about the true nature of the effect. Moreover, if the overall level of true effects is low, then the published literature can end up dominated by false positives because they will continue to be found at 0.05, whereas true positives will be found only at their base rate * power.

What's even more electrifying about Button et al. is a set of meta-anlaysis to determine the degree to which low power is a problem in neuroscience. In imaging studies, power was typically about 8%; in a large set of animals studies between 18-31%, and in papers being meta-analyzed in neuroscience 21%. That's astonishing! This data suggests that most neuroscience studies are not at all adequate for accurate study of the effects of interest. Moreover, there were more poisitive findings reported in these studies than is statistically plausible given their low power.

Overall, Button et al., is a must-read for serious neuroscienctists and should become required reading for graduate programs.

Added an excellent article preview by Howard Eichenbaum. The piece is ostensibly about the finding of spatial-replays in the hippocampus at choice points in a W-maze, and how replays seem to occur for multiple remembered paths, with replays of the correct path predictive of success on the task. More than this, however, Eichenbaum pulls together a range of recent findings to synthesize two traditions of hippocampus research: spatial navigation (O'Keefe & Nadel, hippocampus as a cognitive map) and memory (Squire and Zola-Morgan--hippocampus as a memory processing center). This synthesis depends critically on the ongoing work of Georg Buzsaki, which is conventiently summarized (in part) in an accompmanying perspective on the theta-gamma neural code (which, I must admit, I still don't fully understand). Anyways, I can remember reading Eichenbaum's work as an undergrad which was already drawing parallels between navigation and declarative memory (with a clever olfactory memory paradigm, if I recall correctly)--this paper is an excelletn overview of an emerging and unified view of what exactly the hippocampus might be doing.

Catching up with two papers published in Nature in January (Lee et al, 2013; Volk et al., 2013): both suggest that PKM-zeta is *not* required for LTP maintenance and memory. Both labs independently developed KO mice without PKM-zeta. Both found normal LTP maintenance and normal expression of several forms of LTM. This was true even for LTP maintenance in an inducible knockout, ruling out developmental compensation.

How could previous studies been so wrong? Again, both papers independently confirm that PKM-zip inhibitors are dirty: they can disrupt memory/LTP even in KO mice.

It's nice to see that these results were published in Nature. Maybe this will lead to more concern over specificity of pharmacological manipulations. Well, that's probably asking too much.

Added some new documents related to the billion dollar brain initiatives being launched in the U.S. (maybe) and Europe. The U.S. initiative seems like it will be something like BAM, the Brain Activity Map proposed by Yuste an colleagues (added the Neuron paper in which the project is 'proposed' and a website summary). BAM is intensely focused on multi-unit recording, to record activity in a 'complete neural circuit'. Itt proposes to develop the recording, data processing, and analysis technology required for very large scale recordings (possibly even wirelessly in untethered animals). This is in many ways an exciting goal, but it seems ahead of itself in major ways. First, although focused on 'neural circuits', it's still not clear that we can define whole neural circuits for any complex behavior and in the details they carefully switch to proposals to record from particular areas (like a region of mouse cortex). It seems obvious that a region of mouse cortex doesn't represent a neural circuit for anything--absent co-processing in the thalamus, basal ganglia, etc. The value of just larger scale cortical recordings without any partner structures seems unclear. It's also unclear how we'll extract computational principles if these recordings are made in advance of understanding connectivity.

The human brain project is the European version. They propose generating strategic data (of some sort), developing frameworks and services for collecting/sharing/analyzing data, servicses for large-scale modeling, and services for neurorobotics. Ramp up in 2 years, complete the project in 6! Again, some good ideas (integrating across multiple levels) but seems extremely vague and fantastical. If we want to get seriously systematic, why not adopt the approach biologists have successfully used of focusing on 1 organism and systematically tearing it apart (e.g. e coli, c. elegans, d. melanogaster)? Markram's vision seems to be to somehow organize the chaos that is current neuroscience, but through software tools rather than systematicity in the research endeavors.

My prediction: 2bln wasted.