GEO datasets

[lsand039]

I'm trying to use Jinang's code to score my structures, but I keep running into this error:

Screenshot from 2017-10-05 09-30-37.png (64.03 KiB) Viewed 200675 times

I've specified my structure as "Benestructure.graph.txt" and my data as "2254Sx4G.csv". I'm currently at a loss on how to fix it. I've tried googling for solutions, but I'm unfortunately not familiar enough with python codes.

Below are all the files downloaded from his code and my files.

bayesian_causal_model.git.tar.gz

(652.42 KiB) Downloaded 210 times

[lsand039]

Bene seems to always output a fully connected graph that follows the same order as the variables are listed in the datafile.

Datafiles.xlsx

Sheet number of data corresponds to test number data was used for.

(445.27 KiB) Downloaded 190 times

OrderTest1.png (123.89 KiB) Viewed 200674 times

OrderTest2.png (125.93 KiB) Viewed 200674 times

OrderTest3.png (124.98 KiB) Viewed 200674 times

OrderTest4.png (118.33 KiB) Viewed 200674 times

OrderTest5.png (120.74 KiB) Viewed 200674 times

OrderTest6.png (119.39 KiB) Viewed 200674 times

OrderTest7.png (122.01 KiB) Viewed 200674 times

OrderTest8.png (120.05 KiB) Viewed 200674 times

[lsand039]

Different structures based off Bene-command:
./run.sh: Score: -nan

run.sh4genes.png (123.89 KiB) Viewed 200671 times

./run.sh BDe: Score: -nan

BDe4genes.png (123.89 KiB) Viewed 200671 times

./run.sh BIC: Score : 30.882

BIC4genes.png (17.92 KiB) Viewed 200671 times

./run.sh AIC: Score : 8.000

AIC4genes.png (17.92 KiB) Viewed 200671 times

[lsand039]

I don't have a score for the fully connected Bene structure, but here's the structure for (1): best scoring structure without background knowledge 1st and 2nd degree genes of Alzheimer. This structure was consistent in paths 2, 3, and 5.

(1).png (76.78 KiB) Viewed 200671 times

I ordered the 1st & 2nd degree genes of Alzheimer's according to correlations. The clinical variables were forced to have arcs with Alzheimer's.
Background knowledge files:

(1)mnh.txt

forbidden arcs

(684 Bytes) Downloaded 203 times

(1)mh.txt

Forced arcs

(17 Bytes) Downloaded 203 times

ROC: 0.731618

(1)ROC-AD.png (11.07 KiB) Viewed 200669 times

(1)ROC-nonAD.png (11.06 KiB) Viewed 200669 times

[lsand039]

I don't have a score for the fully connected Bene structure, but here's the structure for (2): best scoring structure with background knowledge 1st and 2nd degree genes of Alzheimer. This structure was consistent in paths 2, 3, and 5.

(2).png (77.53 KiB) Viewed 200670 times

I ordered the 1st & 2nd degree genes of Alzheimer's according to correlations. The clinical variables were forced to have arcs with Alzheimer's.
Background knowledge files:

(2)mh.txt

Forced Arcs

(16 Bytes) Downloaded 205 times

(2)mnh.txt

Forced Arcs

(401 Bytes) Downloaded 198 times

ROC: 0.717182

(2)ROC-nonAD.png (11.09 KiB) Viewed 200669 times

(2)ROC-AD.png (11.04 KiB) Viewed 200669 times

[lsand039]

I don't have a score for the fully connected Bene structure, but here's the structure for (3): best scoring structure without background knowledge 1st and 2nd degree genes of Alzheimer and KEGG genes 4th degree MB with no background knowledge. This structure was consistent in paths 2, 3, and 5.

(3).png (59 KiB) Viewed 200670 times

Background knowledge files:

(3)mnh.txt

Forbidden Arcs

(540 Bytes) Downloaded 188 times

(3)mh.txt

Forced arcs

(16 Bytes) Downloaded 211 times

ROC: 0.73061

(3)ROC-nonAD.png (11.11 KiB) Viewed 200669 times

(3)ROC-AD.png (11.08 KiB) Viewed 200669 times

[lsand039]

I don't have a score for the fully connected Bene structure, but here's the structure for (5): best scoring structure with background knowledge 1st and 2nd degree genes of Alzheimer and KEGG genes 3rd & 4th degree MB with background knowledge. This structure was consistent in paths 2, 3, and 5.

(5).png (88.39 KiB) Viewed 200668 times

Background knowledge files:

(5)mh.txt

forced arcs

(16 Bytes) Downloaded 181 times

(5)mnh.txt

forbidden arcs

(434 Bytes) Downloaded 191 times

ROC: 0.717238

(5)ROC-AD.png (11.07 KiB) Viewed 200668 times

(5)ROC-nonAD.png (11.18 KiB) Viewed 200668 times

[lsand039]

I don't have a score for the fully connected Bene structure, but here's the structure for (6): best scoring structure with background knowledge 1st and 2nd degree genes of Alzheimer KEGG genes 3rd, 4th , &5th degree MB with background knowledge. The BaNJo runs did not agree on a single structure. Here are the scores & log likelihood %:

(6) scores.xlsx

(5.39 KiB) Downloaded 197 times

Here are the top 3 structures. They only accounted for 92% of the log likelihood. There were 2 other structures that would've summed >95%. Please let me know if I should have them posted.
Top scoring structure: 45.89% of total log likelihood score

(6)a.png (104.73 KiB) Viewed 200667 times

ROC: 0.7192

(6)aROC-AD.png (11.2 KiB) Viewed 200666 times

(6)aROC-nonAD.png (11.23 KiB) Viewed 200666 times

2nd top scoring structure: 36.08% of total log likelihood score

(6)b.png (106.25 KiB) Viewed 200667 times

ROC: 0.736538

(6)bROC-AD.png (11.08 KiB) Viewed 200666 times

(6)bROC-nonAD.png (11.08 KiB) Viewed 200666 times

3rd top structure: 9.99% of total log likelihood score

(6)c.png (103.69 KiB) Viewed 200667 times

ROC: 0.736538

(6)cROC-AD.png (11.3 KiB) Viewed 200666 times

(6)cROC-nonAD.png (10.97 KiB) Viewed 200666 times

[lsand039]

Here's the structure for (4): best scoring structure without background knowledge 1st and 2nd degree genes of Alzheimer and KEGG genes 4th&5th degree MB with no background knowledge.
This is the highest scoring structure is below:

(4).png (137.95 KiB) Viewed 200664 times

The scores for the other structure:

(6)scores.xlsx

(5.9 KiB) Downloaded 191 times

Background knowledge files:

(4)mh.txt

forced arcs

(16 Bytes) Downloaded 198 times

(4)mnh.txt

forbidden arcs

(1008 Bytes) Downloaded 193 times

ROC: 0.741332

(4)ROC-AD.png (11.13 KiB) Viewed 200664 times

(4)ROC-nonAD.png (11.16 KiB) Viewed 200664 times

[lsand039]

Here are the arcs that were different in a couple (6)'s top scoring graphs:
CAPN2->MCL1 (not in A)
UQCRC1->MCL1 (not in B,C)

PPP3CB->NFKB2 (not in B)
UQCRC2->NFKB2 (not in A,C)

MCL1->CAPN2 (not in B,C)
Alzheimer->CAPN2 (not in A)

YWHAZ->APBB1 (not in B,C)
SNCA->APBB1 (not in A)

PPP3CB->MYO1F (not in A,B)
ATP2A2->MYO1F (not in C)

PPP3CB->NFKB2; (not in B)
UQCRC2->NFKB2; (not in B)