SMLG (Statistical Machine Learning Group) Discussion Forum

I rework Zhenghua's script so that each BDE score can be analyzed. I also left the values in the negative format. BDE score 7 and 8 are the best network .

##### Rework
> bde=vector()
> bde[1] = -3792.715
> bde[2] = -3681.208
> bde[3] = -3630.236
> bde[4] = -3627.093
> bde[5] = -3615.076
> bde[6]= -3627.093
> bde[7] = -3603.086
> bde[8]= -3603.086
> bde[9]= -3621.687
>
> per1 =vector()
> for(i in 1:9){
+ per1[i] = exp(-log(sum(exp(bde-bde[i]))))
+ }
>
> per1
[1] 2.208718e-83 5.902217e-35 8.088602e-13 1.874395e-11
[5] 3.102972e-06 1.874395e-11 4.999984e-01 4.999984e-01
[9] 4.175005e-09

> ##### Rework
> bde=vector()
> bde[1] = -3792.715 ###1000 restarts
> bde[2] = -3681.208 ###5000 restarts
> bde[3] = -3630.236 ###10000 restarts
> bde[4] = -3627.093 ###20000 restarts
> bde[5] = -3615.076 ###20000 restarts
> bde[6]= -3627.093 ###20000 restarts
> bde[7] = -3603.086 ###40000 restarts using best DAG from 20000
> bde[8]= -3603.086 ###40000 restarts using best DAG from 20000
> bde[9]= -3621.687 ###40000 restarts random DAG
> bde[10]= -3598.474 ####80000 restarts using best DAG from 40000
> bde[11]= -3596.369 ####80000 restarts using best DAG from 40000
> bde[12]= -3596.341 ###160000 restarts using best DAG from 40000
>
>
>
> per1 =vector()
> for(i in 1:12){
+ per1[i] = exp(-log(sum(exp(bde-bde[i]))))
+ }
>
>
> per1
[1] 2.483361e-86 6.636127e-38 9.094378e-16 2.107466e-14
[5] 3.488810e-09 2.107466e-14 5.621707e-04 5.621707e-04
[9] 4.694145e-12 5.660233e-02 4.645412e-01 4.777321e-01

Pearson Correlation's

Mouse dataset

Mouse network dot file
Mouse netfile

Human network dot file, netfile, data

Rat network dot file
Rat network net file ( I had to save it as a text file to post)
Data for network

Data for BANJO

Here are the unformatted networks for 3 independent 1-hour banjo runs on the BANJO12 data (552 mouse genes+fasting).
Fasting is highlighted in yellow - ctrl-F to find the node does not work in Firefox, but works in Edge and Chrome.

The mouse representative network (roughly 30 genes) -
I used the genes from (1) the MB of fasting, (2) the genes of interest (3) top 10 genes with highest probabilities 'finalmouseprobs8_8' and (4) the MB of the top 10 genes.