I *THINK* g_rms has no mpi option as of now. So I think, it wouldn't be possible for you to carry out analysis over pbs script. I would rather recommend you to download the files to the local machine and carry out the analysis

Dear Sir

I don't think you will have problem with g_rms for bigger system. I'm working with ~3 lakhs atom size systems and also running microseconds timescale. I do rmsd plot in my local machine 4core, 8gb RAM. g_rms don't need huge computation. For rms matrix you may need ~16gb RAM. Still if you want to run g_rms with pbs then just copy paste in your pbs script the same command you are using in local machine. 1 processor is enough to calculate huge trajectory and in general server's compute node RAM is enough. If you have any problem, then share the pbs script, I'll fix it.

Thank you

Dear Sir

I'm not agree with Thirumal Sir. No need to download in local machine. You should do all analysis in your compute node or high memory node specialised for trajectory analysis. As local machine won't be able to deal with huge trajectory and analysis like pca or normal mode.

Dear krishnendu

Most of the clusters never allow you to submit your job directly on clusters (eg. g_rms on qlogin). When you prepare .tpr file on local system, then analysis also should be possible with local system. You can't wait for the pbs script to run in queue for rmsd or other analysis as minimal processors is sufficient. Rather waiting in queue and wasting the time, you can go with analysis in local system. You also can use GUI like VMD to analysis in local system, which is not possible in clusters

Dear Sir

@Thirumal; I agreed with you that most of the clusters won't allow you to submit job directly and it's a bad idea to do analysis in head node. I told you to do analysis in compute node. You can ssh to particular compute node you want and do your analysis there.  Previously I've also written that for RMSD single processor is enough and no need of multiprocessor for RMSD. But for RMSD Matrix you may need huge computation for microsecond level trajectory. We specially design particular node for analysis, which is mostly called high memory node and people prepare it with high RAM (~512gb) for the analysis. Using VMD in GUI for analysis is a bad idea for big trajectories. You can't do big trajectory analysis in GUI. One should use VMD in text mode and use tcl script for big trajectory analysis by generating small chunks for trajectories (5ns-10ns) and do analysis in using loop in script for microsecond level simulation. 

But being a beginner, he can use GUI and no wrong in that. But command it's always best

But as he mentioned his trajectory is huge so it's not good idea to recommend GUI for analysis. And he has 6 papers in bioinformatics, which suggesting he's not beginner and he can do it and these scripting are very easy and most of these scripts are available in net.

@Krishnedu

But you can't expect VMD installed on all the clusters (binaries never support all the time). You can install VMD on your local system.

Recently I have installed vmd at IISER Kolkat HPC without root permission. You can install vmd in your home in any HPC. You don't have to worry for installation. 

Thanks a lot Krishnendu and Thirumal for useful suggestions. My file was of 98 GB and local machine have 24 GB RAM, that is why system was unable to load the file. Anyways, i am working on a sub script for main pbs and hopefully will generate the plot soon. Moreover, can you help me in protein-DNA simulation? (in free time if possible)

I would suggest you use the latest Charmm forcefield for protein-DNA simulations.