Biomedical Engineering Reference
In-Depth Information
11. Comparison
of Functional
Interactions with
Multiple
Techniques
We have presented an array of techniques for assessing functional
interactions in this manuscript. Each technique we have presented
describes a different aspect of how neurons interact, and provides
a distinct perspective to neurons' functional interactions. To pro-
vide a multivariate portrait of functional interactions, we present
a table describing the interactions between two pairs of neurons
(
Table 7.1
).
For the first pair of neurons, (sig033b dmPFC &
sig038b dmPFC), we find similar response properties (holding
the lever down), strong cross-correlations (the ratio of peaks to
peaks in random noise is shown), suggesting a synaptic relation-
ship. We also find strong rate and trial-by-trial rate correlations;
however, we find no instance of JPSTH interactions, predictive
interactions, and hence, no network interactions. This type of
pair, while interesting to many neuroscientists because it describes
a monosynaptically connected pair of neurons at least 250 um
distant from each other, has no correlations that change in rela-
tion to stimulus presentation and is poorly predictive of reaction
time. This is a classic example of a
noise
correlation
(38)
,thatis,
despite the fact that these neurons themselves are having similar
behavioral modulations, the communication between this pair of
neurons is not directly related to behavior. This pair belies the
importance of our approach of comparing correlative techniques,
as no one measure of functional interactions provides a complete
picture of neuronal interactions.
Table 7.1
Functional interactions
Chance
Pair
Pair
-
sig033b dmPFC
sig040a dmPFC
sig038b dmPFC
sig024a MC
Principal component analysis
9.8
Hold component (PC 2)
No common component
Cross-correlation
1.1
12
1.0
JPSTH
0.2
0
0.35
Rate correlation
0.14
0.91
-0.03
Trial-by-trial correlation
0.18
0.83
0.17
Predictive information
∼
0.72
0.83
Redundancy
-0.2
-0.002
-0.46
