GEO datasets

[lsand039]

After another 48 hr BaNJO run on Paths 2,3, and 4, the Path 4 Hour 48 graph was the new top scoring graph which was significantly better (100%) than the 2nd top scoring graph, the first 48 hr BaNJO run at Path 3.

Best Scores.xlsx

(7.35 KiB) Downloaded 170 times

MB1 genes:
GLRX5
TRIP10

MB2 genes:
CACNB2
CLTA
HMG20B
SLC25A1
IFT57
PSME3
TIMM8B
UCHL1
STX12
OSBPL3
PCSK1
FGF14
MAPK6

APP, PSEN1, and PSEN2 showed up in the 6th degree MB; APOE showed up in the 7th degree MB. Below are images of the full structure, the 1st & 2nd degree MB, and a structure that shows the genes separating Alzheimer's from APOE, APP, PSEN1, and PSEN2.

P4H48 images.tar.gz

(1.55 MiB) Downloaded 161 times

[lsand039]

This image is the best scoring graph and has the genes in the KEGG Alzheimer's pathway highlighted in teal. There are 171 different genes within the KEGG pathway, and 107 are among the 8092 genes common to all our datasets. None of the 1st or 2nd degree MB blanket genes were part of the KEGG Alzheimer's pathway.

Hour48-4KEGG.dot.svg.tar.gz

(1.55 MiB) Downloaded 152 times

[lsand039]

Here's the file calculating the correlation of genes and clinical variables to Alzheimer's using Z-scores.

8092Genesx2254SamplesCorrelations.xlsx.tar.gz

(259.84 MiB) Downloaded 168 times

[lsand039]

The last 48 BaNJO run just finished, and the top scoring graph is still from Path-4 at 48 Hours. I'll be working on finding the MB degree for the KEGG Alzheimer genes.

Below is the score comparison of all the trials.

Best Scores.xlsx

(7.47 KiB) Downloaded 155 times

[lsand039]

Here is an ordering of the KEGG and MB genes I've already found in relation to Alzheimer based off the data.

P4H48KEGGorder.xdsl

(19.32 KiB) Downloaded 168 times

Here I tried to merge the order from the KEGG pathway and the order from the data. I used the ordering from the data to order genes that have the same order on KEGG.

KEGGorderMB5.xdsl

(4.34 MiB) Downloaded 170 times

Some of the genes' order from the data conflicted with the ordering from KEGG. I noticed the genes linked to Ca2+ had a lot of issues. Once I have the full ordering of all the KEGG genes from the data, I can try to better reconcile the KEGG and data ordering.

[lsand039]

I've been running BaNJO on the data using the background knowledge of the confounding structure I used in my thesis.
In the file below, I set Age, Sex, and Brain Region as the highest order (no variable can be parents to these variables) and Alzheimer's as the lowest order (Alzheimer's cannot be a parent to any variable). I kept the order APOE>APP>PSEN1>PSEN2 as background knowledge by specifying the following:
APOE cannot have APP, PSEN1, and PSEN2 as parents.
APP cannot have PSEN1 and PSEN2 as parents .
PSEN1 cannot have PSEN2 as a parent.

CS_order.txt

(185.44 KiB) Downloaded 173 times

For the 1 hour, 2 hours, and 4 hour runs, using the background knowledge produced better scores than the BaNJO runs without background knowledge. This was not true after 8 and 12 hours, though scores got better as time went on within the BaNJO runs using background knowledge. So far the 2nd 48 Hour BaNJO run on Path 4 has the all time highest score. Here are the scores so far.

Best Scores.xlsx

(7.97 KiB) Downloaded 170 times

I'll be updating results once the 48 hour run is complete.

[lsand039]

The 48 hour and 36 hour BaNJO run with background knowledge did not produce better scores than the runs without background knowledge. When only looking at the background knowledge scores, the 48 hour run on Path 4 had the best score (100%).

Best Scores.xlsx

(8.41 KiB) Downloaded 153 times

.

I'm currently running a second background knowledge BaNJO run using the KEGG ordering. I'll be posting results once the 12 hour run is complete.

[lsand039]

Using background knowledge from KEGG produces better scores than the BaNJO runs with no background knowledge. Here are the scores so far:

Best Scores.xlsx

(8.51 KiB) Downloaded 184 times

Since the 12 hour score produced the best score, I'm doubling the time used on BaNJO. I have a 24 hour run then a 48 hour run going on Paths 2, 3, and 4.

[lsand039]

I finished finding all the KEGG genes' MB by hand. I'll check my results with Efrain's code once it's finished. 2 of the KEGG genes (SDHC and MME) were not connected to the structure at all.

Here's the image and the dot text file with the KEGG genes connected to Alzheimer's:

FreeRun_KEGG.dot.svg.tar.gz

(340.44 KiB) Downloaded 166 times

FreeRun_KEGG.dot.svg.tar.gz

(340.44 KiB) Downloaded 166 times

This file will show you the path for each KEGG gene to AD.

FreeRun_KEGG.dot.svg.tar.gz

(340.44 KiB) Downloaded 166 times

Note that these are for the BaNJO results with no background knowledge.

[lsand039]

The 24 and 48 Hour BaNJO runs with the KEGG order as background knowledge is complete. The best score was from 48 hours on Path-2 (100% of scores).
MB1 genes:
MCL1
YWHAZ
MB2 genes:
LRRN3
MYO1F
CYP4F12
DHRS3
BMP7
DVL2
NFKB2
SMU1
WSB2
GRIA1
SRRM2

This, however, is still significantly lower than the best score with no background knowledge (48 hours on Path-4, Trial 2).

Best Scores.xlsx

(8.63 KiB) Downloaded 172 times