I strongly recommend you using Multiwfn program (http://sobereva.com/multiwfn) to carry out molecular orbital composition analysis, see Section 4.8 of its manual for examples. The use is very easy, and you can directly use .fch file produced by Gaussian as input file.

Multiwfn supports numerous ways to calculate orbital composition, including Mulliken, SCPA, Stout-Politzer, Hirshfeld, Becke, Hirshfeld-I and NAO methods. If you only need atomic contribution to MOs, Hirshfeld or Becke method is commonly the best choice.

PS: I wrote a paper many year ago detailedly discussed and compared various methods for evaluating molecular orbital composition, see http://sioc-journal.cn/Jwk_hxxb/EN/abstract/abstract340458.shtml (in Chinese). As illustrated in this paper, the popular Mulliken analysis, can be employed to calculate atomic contributions to MOs; however, the result will be meaningless when diffuse functions are used, while the Hirshfeld and Becke methods mentioned above do not have this drawback.

if you use keyword Pop=Reg, the MO coefficients will be included in your output.  

I would be careful with things like Pop=reg. Mulliken populations are vague at best and basis-set-dependent-nonsense at worst. The first thing to keep in mind is that canonical MOs aren't the nice clean ones that everyone likes to think in, they're in principle diffused across the entire molecule. In order to be able to marry our nice, accurate calculations with our messy, comfortable heuristics, we have to come up with some scheme or basis set rotation that will localize our one particle spacial orbitals in places that we expect and are useful. You used one already when you performed Natural Orbital analysis. IN THAT SCHEME you can read off things that attribute charges to particular atoms. However, using Bader's AIM, Vornoi, Hirshfeld or various electrostatic methods, you can get different answers that are mathematically equivalent (in the sense that one linear transformation is as valid as another). If you have a sense of what you want to get out of this calculation, look in the Gaussian output file and search for "NBO"; you should be able to find a listing of all the atoms with charges and which other atoms they have "natural bonds" with. If you're not sure (or what the hell, just for fun) then you should read something like the following: Comput Chem 25: 189–210, 2004. It gives a rundown of why Mulliken shouldn't be trusted (I've reviewed many a paper that presents Mulliken charges and I always ask for another method or a removal of the partial charge based argument) and gives an overview of what you can expect from the other,  more sophisticated methods. 

Enjoy!

I strongly recommend you using Multiwfn program (http://sobereva.com/multiwfn) to carry out molecular orbital composition analysis, see Section 4.8 of its manual for examples. The use is very easy, and you can directly use .fch file produced by Gaussian as input file.

Multiwfn supports numerous ways to calculate orbital composition, including Mulliken, SCPA, Stout-Politzer, Hirshfeld, Becke, Hirshfeld-I and NAO methods. If you only need atomic contribution to MOs, Hirshfeld or Becke method is commonly the best choice.

PS: I wrote a paper many year ago detailedly discussed and compared various methods for evaluating molecular orbital composition, see http://sioc-journal.cn/Jwk_hxxb/EN/abstract/abstract340458.shtml (in Chinese). As illustrated in this paper, the popular Mulliken analysis, can be employed to calculate atomic contributions to MOs; however, the result will be meaningless when diffuse functions are used, while the Hirshfeld and Becke methods mentioned above do not have this drawback.

Thank you, Michael Rubin, Adam W. Pelzer,  Tian Lu for your valuable suggestions.

 J Mol Graph Model. 2008 Oct;27(3):275-85. doi: 10.1016/j.jmgm.2008.05.002. 

In the above-mentioned article, they have mentioned that Molecular orbital (MO) HOMO (highest occupied molecular orbital) and LUMO were carried out for N1, N3, and AG7088 by PM3MM (N1 only) and B3LYP/6-31G(d).

How to do such calculations and get the plot for HOMO and LUMO orbitals?

The name 'Molecular Orbital' itself says it is orbital on molecule not atomic. You can't get Molecular orbital for a particular atom. Please read carefully the paper which you have referred. The N1, N3 here are not  Nitrogen atoms.  They named  two of their system as N1 and N3. Hope this will help.

thank you, Debasish Mandal

The programs QMForge-2.1 or Multiwfn are very good if you are looking for the charge decomposition analysis, otherwise you can use the following route

#P B3LYP/6-31G(d)  opt freq # SCF(maxcyc=100) gfinput gfprint pop=full