ALLEN BRAIN ATLAS ADULT HUMAN Whole brain microarrays Agilent 8x60k array starting from 4x44k Agilent Whole Human Genome probe set 2 probes for 93 of genes 21k unique Entrez Ids ID: 582744
Download Presentation The PPT/PDF document "http://www.brain-map.org" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
http://www.brain-map.orgSlide2Slide3
ALLEN BRAIN ATLAS: ADULT HUMAN
“Whole brain” microarrays:
Agilent 8x60k array
, starting from 4x44k Agilent Whole Human Genome probe set
2+ probes for 93% of genes [~21k unique
Entrez
Ids]Slide4
Gene FinderUser navigates to voxel
-
of-interest in reference
atlas volume and
a
fixed threshold
AGEA correlation map appearsGet a gene list from ABA is returned. Slide5
AGEA Gene Finder Tool enables users to search a local anatomic region of interest for genes that exhibit localized
enrichment
Finding
genes
with
highly localized expression
is of
neuroscientific interest - structural relationships, evidence for refinement of structural boundaries.Slide6Slide7
For seed s, correlation value t, find set of voxels N(t,s
)
Let
B(s
) =
N(T,s
)
Let
A(s
) be local neighborhood of highest correlated
voxels
The Finder AlgorithmSlide8
Ranked List of Genes
Computation is independent for 16 brain regions
R
with unique intra‐correlation patterns
Regions include - cortex, hippocampus, striatum, thalamus, olfactory bulb, cerebellar cortex, hypothalamus, midbrain and hindbrain.
Special Regions - Ventricular areas, medial
habenula
,
caudoputamen
, deep cortical layers, olfactory nerve layer of the olfactory bulb,
zona
incerta
and
inferiorSlide9
Cortical Map
Genes in
superficial layers
have sharp drop
in
correlation depth-wise
Transition not smooth – L5 & L6:
column a
Vice-versa;
E
xpression in deep layers reduces correlation in superficial layers
L
aminar effects -
seeds
in
somatosensory L6 have lower L4 correlation (
column
d
) than seeds
in L2
/3Slide10
Visualizing Correlations
A
llows interpretation of relative
correlations across layers and regions.
M
ean
correlation
is highest in
the domain containing the seed
Use representation to determine
dominant
area (columns) or layer (rows) to show that adjacent
layers
have positive expression
correlation
S
trongest concordance between
L5 and
L6
N
on
-adjacent
layers
-
negative
correlation with
anatomic
proximity: physically
distant layers
less
likely to exhibit gene
coexpression
.Slide11
Multi-dimensional Scaling
D
omain
-to-domain
correlations as
measure
of similarity
D
ata is visualized by multidimensional scaling (MDS)Clustering method Distance between
points (
domains) is proportional to their correlation
MDS recapitulates
the
basic laminar and areal relationships of the
neocortex
Proximal
and functional relationship of
SSp
and
SSs
L
ower
concordance of VISp with other regions.Slide12
Multi Dimensional ScalingSlide13
Multidimensional Scaling
From a
matrix of
distances…
Kruskal & Wish, 1978Slide14
MDS
…it
calculates
a map…Slide15
MDS
What does the MDS algorithm do?
…but it
cannot tell
the
orientation
and the
meaning of
the axes.
Tuesday, May 5, 2009Slide16
MDS
Shepard, 1963:
•
Morse-codes presented in pairs to naïve observers (each possible
combination)
•
Task - Same
/
different
•
Confusion matrix (% same responses): can be interpreted as a
dissimilarity matrixSlide17Slide18Slide19
MDS Algorithm
•
Given a set of similarities (or distances) between every pair
of N items
•
Find a representation of the items in few dimensions
•
Inter-item proximities “nearly” match the original similarities
(or distances)
65
Tuesday, May 5, 2009Slide20
ObjectiveSlide21
Kruskal’s StressSlide22
Next