How to run job on ijcluster:  	
   
   -- interactive mode:
   *********************
Hi,
You can have a compute node with lots of memory and 20 cores all to yourself.From ijcluster, just run 
 srun --cpus-per-task 20 --pty bash -i
and you will get a prompt on a compute node reserved for you until you exit.
Christian   

   -- on Node by srun interactively:
   **********************
login to ijcluster 

-- use "srun" to run the job ===============

cd /home/USER_NAME/PP_Docking/MEGADOCK/megadock-4.1.1/your_folder

srun --cpus-per-task 20 --pty bash -i 
../megadock -R Vimentin.pdb -L Annexin_A2.pdb -o vim_ann.out -O > log.out 2>&1 &



   -- on Node by srun NON-interactively:
   **********************
login to ijcluster 

-- use "srun" to run the job ===============

cd /home/USER_NAME/PP_Docking/MEGADOCK/megadock-4.1.1/_Vim_Ann
srun --cpus-per-task 20 -o run.out -e run.out ../megadock -R Vimentin.pdb -L Annexin_A2.pdb -o vim_ann.out -O &



squeue --user USER_NAME --states all

-- clean-up ============
-- manually kill job
Ctrl-C 
- or (find jobs)  
ps -ef | grep USER_NAME
- and (kill jobs)
kill -9 


-- free allocated node (the job will continue to run) ==================
- find allocated (be salloc) JOBID 
squeue 
- cancel JOBID
scancel JOBID
- or If you are in salloced-bash-shell the just type: 
exit
   
   
   
   
   
	
	How to use MEGADOCK for PPI prediction
	=======================================
https://www.bi.cs.titech.ac.jp/megadock/ppi.html   (use Chrome to translate Japanise to Englist)

	<1>	Docking using MEGADOCK
- First, use MEGADOCK to calculate the docking of the two PDBs whose interaction you want to predict. 
- At this time, please set the option -t 3 -N 10800


cd /home/USER_NAME/PP_Docking/MEGADOCK/megadock-4.1.1/_REC_LIG
srun --cpus-per-task 20 -o run.out -e run.out ../megadock -R rec.pdb -L lig.pdb -o dock.out -t 3 -N 10800
nohup srun --cpus-per-task 20 -o run.out -e run.out ../megadock -R rec.pdb -L lig.pdb -o dock.out -t 3 -N 10800 &	  
  
   
	<2>	Improving accuracy by reranking (optional)   
- I skipped it for now.  

	<3>	PPI evaluation value (E) using the PPI evaluation value calculation and evaluation tool.
a) When reranking is not performed (Note: I use this because I skipped reranking)
ppiscore

../ppiscore dock.out 10800

A threshold value (E*) is set for the calculated PPI evaluation value (E), and the presence or absence of PPI is determined based on the magnitude of the value with (E*).

E > E*   → interact
E <= E*  → do not interact.


	Note by MA: Further on this page the proved in what (E*) interval the above works the best
Sensitivity, recall and selectivity (precision): 
- It is usually set to around E*=6.0~7.0. 
- If you want to increase the selectivity, increase it to about E*=10.0~13.0. 
Regarding the change in accuracy due to (E*), please see the results verified using the complex of ZLAB docking benchmark ( Mintseris et al. , Proteins 2010 ) at the bottom.
(So, I better read this section on some point !!!)