Read on Twitter
If you read the recent NHP optogenetics paper, you might have concluded that opto works well in monkeys & we should be using it to speed discovery in basic, translational, & clinical neuroscience.
Not so fast, say @vinnycosta_phd, @markgbaxter and I.
https://www.biorxiv.org/content/10.1101/2020.12.10.420331v1
1/
Not so fast, say @vinnycosta_phd, @markgbaxter and I.
https://www.biorxiv.org/content/10.1101/2020.12.10.420331v1
1/
First, we want to thank the original paper's team for making public a resource that includes successes & failures & unpublished data. NHP meta/open science efforts are rare and SUPER important b/c of ethical & practical constraints of NHP science.
https://www.sciencedirect.com/science/article/abs/pii/S0896627320307510
2/
https://www.sciencedirect.com/science/article/abs/pii/S0896627320307510
2/
Original analyses in the paper were at the level of individual injections (based on rodent lit). It's not clear that's meaningful in the primate brain, so we reanalyzed the data nesting injections w/in monkeys/brain areas resulting in a unit of analysis we call "experiments".
3/
3/
The original paper reports opto is highly successful (91% successful including weak effects; 76% successful including only strong effects), but that claim is made based on an effect on physiology OR histology OR behavior. Very few papers actually measure behavior.
4/
4/
This is an issue in light of claims (made in the paper) about the usefulness of opto for speeding clinical neuroscience on the basis of these "functional" outcomes.
5/
5/
Behavior+physiology or behavior+histology success rates were lower: 62.5% w/weak effects; 48.6% & 57.5% strong effects. Most experiments did not include a behavioral readout at all (309 of 383).
6/
6/
It's important to point out that most of the data points come from experiments on primary sensory or motor cortex where it should be relatively straightforward to modulate stereotyped behavior.
7/
7/
We detail other issues (including how on earth to get light to deep structures without doing other damage?) as well.
Our take home is that opto in monkeys isn't ready for primetime right now and our major concern is that the original paper will be leveraged to claim it is.
8/
Our take home is that opto in monkeys isn't ready for primetime right now and our major concern is that the original paper will be leveraged to claim it is.
8/
This has real consequences for the direction science heads. We have all experienced NIH folks telling us that they will not fund classic methods b/c we should be using optogenetics.
9/
9/
There's no doubt that tool development is a worthy cause, but in our opinion not at the expense of pursuing hypothesis-driven questions in the service of speeding translational and clinical neuroscience.
10/
10/
Continuing to use classic methods (e.g., electrical microstimulation, lesions, pharmacology, etc.) to answer hypothesis driven questions while also developing new tools is the way to go, in our opinion.
11/
11/
Not everyone has the resources to do tool development or risk implementing new tools that may not bear fruit, and progress is delayed by expending energy on tool development to address questions that could be answered as well or better by less "fancy" methods.
12/
12/
TL;DR: a reanalysis of the open data resource suggests that when we group injections by monkey/brain area, & code outcomes to include behavior, the chance of success of NHP opto is about a coin flip. We need to be aware of this as we make choices about what science is done. fin/
