GEO datasets

[lsand039]

Here is the Methods section of the work I've done so far. I tried to combine what we've been doing over the summer with results we got last year regarding the order of APOE, APP, PSEN1, and PSEN2 and their relationships with the confounding variables.

I've added some questions in the comments and this draft is open for feedback.

[lsand039]

I've redone the Bene Structure on the 4 genes of interest using our new dataset. Bene wouldn't score the structure, I scored it on BaNJo.

Benestructure.graph.2017.09.19.09.32.46.png (139.09 KiB) Viewed 88824 times

99.9999999992724 % of total score
Here are the setting files I used for this structure:
Banjo Input files

settingsBene.txt

settings file

(5.8 KiB) Downloaded 170 times

BeneStructure.txt

Specified structure

(95 Bytes) Downloaded 154 times

4genes.txt

Data

(35.28 KiB) Downloaded 172 times

Here's the confounding structure I scored on BaNJo.

CS.graph.2017.09.19.09.46.56.png (112.58 KiB) Viewed 88824 times

6.78572483028301e-10 % of total score
Here are the setting files I used for this structure:

settingsCS.txt

Settings file for Confounding Structure

(5.79 KiB) Downloaded 174 times

CS.txt

Specified Confounding Structure relationships

(81 Bytes) Downloaded 171 times

CSmnh.txt

Specified forbidden relationships

(61 Bytes) Downloaded 176 times

Min. Significance Value: 2.60e-6
Max. Significance Value: 1.61e-3

Distributions:

Distributions.png (13.63 KiB) Viewed 88823 times

ROC
Alzheimer's

ADROC.png (9.61 KiB) Viewed 88823 times

[lsand039]

ROC non Alzheimer

nonADROC.png (9.6 KiB) Viewed 88823 times

[lsand039]

The tab named "Correlation Rankings" show the Correlation Values for the KEGG genes to Alzheimer's .

The lowest correlated KEGG gene was IL1B (0.0024) and the highest was NDUFA9 (0.34).

It looks like the KEGG gene correlations are all over the place. The genes range from 52nd to 7969th highest correlated gene out of 8092 genes.

[lsand039]

This contains the comparisons of MB degrees between the structures with and without Background knowledge. I've only searched up to the 8th degree MB, but some of the genes listed as 8th degree in the KEGG background knowledge may actually have a higher MB degree. I'm currently checking those up to the 11th MB degree.

MBComparison.xlsx

(8.12 KiB) Downloaded 181 times

Only 15 of the genes had the same MB in both structures. 42 of the genes may have a MB degree higher than 8 in the KEGG background knowledge structure. 48 of these genes had completely different MB degrees.

[lsand039]

The ROC for the 1MB structure without background knowledge:

MB1 for no Background Knowledge For AD

MB1NoBKAD.png (9.77 KiB) Viewed 88818 times

MB1 for no Background Knowledge For Non-AD

MB1NoBKnonAD.png (9.78 KiB) Viewed 88818 times

MB1 for no Background Knowledge

MB1NoBK.png (6.93 KiB) Viewed 88818 times

The ROC for the 1MB structure with KEGG background knowledge:

MB1 for KEGG Background Knowledge For AD

MB1BKKEGGAD.png (9.81 KiB) Viewed 88818 times

MB1 for KEGG Background Knowledge For Non-AD

MB1KEGGBKnonAD.png (9.92 KiB) Viewed 88818 times

MB1 for KEGG Background Knowledge

1MBKKEGG.png (6.82 KiB) Viewed 88818 times

[lsand039]

Here's an update on the MB degree comparisons between the structures with and without background knowledge:

MBComparison.xlsx

(9.14 KiB) Downloaded 154 times

Highest MB degree listed that is accurate: 11
KEGG genes with the same MB degree: 19
KEGG genes found later with background knowledge: 54
KEGG genes found earlier with background knowledge: 28
KEGG genes with ±1 MB difference: 59
KEGG genes with ±2 MB difference: 82
Total KEGG genes: 107

Genes not connected in structure without background knowledge: MME, SDHC
Genes not connected in structure with background knowledge: FADD, IL1B

Still need to find MB degree in structure without background knowledge: CAPN1
Still need to check MB degree is higher than 11 in structure with background knowledge:CALML3

[cwyoo]

Here's an update on the MB degree comparisons between the structures with and without background knowledge:

MBComparison.xlsx

Highest MB degree listed that is accurate: 11
KEGG genes with the same MB degree: 20
KEGG genes found later with background knowledge: 54
KEGG genes found earlier with background knowledge: 28
KEGG genes with ±1 MB difference: 58
KEGG genes with ±2 MB difference: 82
Total KEGG genes: 107

Genes not connected in structure without background knowledge: MME, SDHC
Genes not connected in structure with background knowledge: FADD, IL1B

Still need to find MB degree in structure without background knowledge: CAPN1
Still need to check MB degree is higher than 11 in structure with background knowledge: CACNA1F, CALML3

Please create the following datasets (add clinical variables to all datasets) and learn BN with Bene (if Bene does not run, run it with Banjo) and report ROC and AUROC:
(1) Dataset with following genes (best scoring structure without background knowledge 1st and 2nd degree genes of Alzheimer):
GLRX5
TRIP10
CACNB2
CLTA
HMG20B
SLC25A1
IFT57
PSME3
TIMM8B
UCHL1
STX12
OSBPL3
PCSK1
FGF14
MAPK6

(2) Dataset with following genes (best scoring structure with background knowledge 1st and 2nd degree genes of Alzheimer):
MCL1
YWHAZ
LRRN3
MYO1F
CYP4F12
DHRS3
BMP7
DVL2
NFKB2
SMU1
WSB2
GRIA1
SRRM2

(3) Add the following genes to (1) (KEGG genes 4th degree MB with no background knowledge):
GRIN2A
NDUFB3
UQCRQ

(4) Add the following genes to (1) (KEGG genes 4th and 5th degrees MB with no background knowledge):
GRIN2A
NDUFB3
UQCRQ
CACNA1F
CASP9
MAPT
NDUFA10
GNAQ
GNAQ
NDUFB6
COX5A
COX7B
MAPK1
NDUFB5
NDUFC1
RTN3
SDHB
BACE1
GRIN1
NDUFA9
PLCB1
IL1B
UQCRC1
NDUFA5
UQCRC2

(5) Add the following genes to (2) (KEGG genes 3rd and 4th degrees MB with background knowledge):
GRIN2D
NDUFA5
UQCRC2
SNCA

(6) Add the following genes to (2) (KEGG genes 3rd, 4th and 5th degrees MB with background knowledge):
GRIN2D
NDUFA5
UQCRC2
SNCA
UQCRQ
UQCRC1
ATP2A2
CAPN2
APBB1
PPP3CB
CYC1
NCSTN
PLCB4

[lsand039]

Here is the structure from Bene for (1).

1.MB1_2NoBK.png (1.83 MiB) Viewed 88812 times

Data file:

(1)MB1&2NoBK.txt

(85.97 KiB) Downloaded 180 times

I'm working on getting the ROC, but GeNIe keeps crashing as I add arcs to the structure. It's gotten to the point where GeNIe crashes every time I open the file. I just need to add the arcs where GLRX5 is the parent.

1.MB1_2NoBK.xdsl.tar.gz

(1.62 MiB) Downloaded 177 times

Here is the structure from Bene for (2)

2.MB1_2BK.png (1.26 MiB) Viewed 88806 times

Data file:

(2)MB1&2BKKEGG.txt

(77.15 KiB) Downloaded 171 times

This is the structure from Bene for (3)

3.MB1_2_4NoBK.png (2.98 MiB) Viewed 88810 times

Data file:

(3)MB1_2_4NoBK.txt

(99.2 KiB) Downloaded 161 times

The is the structure from Bene for (5)

5.MB1_2_3_4BK.png (2.54 MiB) Viewed 88806 times

Data file:

(5)MB1_2_3_4BKKEGG.txt

(94.78 KiB) Downloaded 170 times

[lsand039]

I was having issues with Bene giving different orders of variables when the data file has the same variables listed in different orders. I can confirm that Bene gives equivalent structures if the order is not the same:

BeneStructureTest.graph.2017.10.04.11.34.25.png (111.94 KiB) Viewed 88806 times