We hypothesized that cadherin expression was key. So we used BARseq2 to detect cadherin expression and cell type markers in auditory and motor cortex
(4/n)
Laminar patterns of gene expression obtained by BARseq2 were similar to the that seen in the Allen Brain Atlas
(5/n)
The sensitivity of BARseq2 was similar to 10x v3 in this particular experiment, but if we push it we can reach at least 60% of RNAscope
(6/n)
We also determined transcriptomic cell types by detecting additional cell type markers in the motor cortex.
7/n
Combining in situ sequencing of barcodes and endogenous genes, we correlated cadherin expression to projections in 1,349 single neurons in auditory and motor cortex.
Gene expression+single neuron axon projection patterns in 1000 neurons, from just a handful of animals!
8/n
Our data recapitulated known differences in gene expression and projections between neuronal classes, types, and subtypes.
For example, neurons that project only ipsilaterally were mostly IT4 transcriptomic type (i.e. L6 Car3 type)
9/n
How do these cadherins relate to transcriptomic cell types?
To find out, we grouped cadherins into co-expression modules and found that these modules were associated with multiple branches of the transcriptomic taxonomy
10/n
Our results suggest that the relationship between gene expression and axonal projections in the cortex is complex
Probably not fully described by the transcriptomic types alone
Maybe the key is to look during development? Or at other genes? Stay tuned...
11/11
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Biological neural networks (BNNs) are much more energy efficient than artificial NNs. The human brain uses about 15-20W, whereas eg training a big ANN causes the lights to dim in Boston for a day or 2.
Why?
1/n
The usual explanation for the energy efficiency of BNNs is sparse spiking.
Eg firing rate in the cortex is 1-0.1 Hz or less. Most neurons are silent most of the time.
Sparse spiking saves energy bcs repolarizing neurons requires energy expensive Na+/K+-ATPase pump
2/n
And we have some specializations related to dexterity (which enables tool use) and bipedality.
And our theory of mind is really sophisticated. We primates are really good at predicting others' actions; and when we started to predict our own actions, consciousness emerged
2/n
I think every (or maybe just many/most) species have specializations. Eg bats are specialized for ultrasound localization; elephants surely have brain specializations for their trunks; etc.
3/n
We touched on a lot of interesting subjects at the great Salon yesterday
But i would like to dig into one where i think we failed to communicate: what it means to be "close" in genetic space, and why i think it's relevant. (1/n)
Suppose i am trying to solve eg a prediction task, where I take inputs from a set X and try to predict Y. Maybe X is a set of images, and Y is a set of labels.
Now let's say that i try to solve it with my favorite algorithm, and fail. (2/n)
I take it to one of you, and you say, "Oh, i see the problem, you just need to pre-process your images with this edge detector" or "You just need to use weight momentum to make your network converge". (3/n)