demo(nma) suggestion

[Barry Grant]

Not sure that it is a good idea to download all the kinase PDB files during this demo - It could be annoying for a user to have these appear on their disc without warning.

Suggest using the transducin data and plot as appear later in vignette...

[Lars Skjærven]

Hmm.. I was just thinking to calculate NMA on 53 structures could also be a bit annoying.. how about:

tmp.dir <- tempdir()

### Download a set of Kinase structures
ids <- c("4b7t_A", "2exm_A", "1opj_A", 
         "4jaj_A", "1a9u_A", "1tki_A")

### Download and split by chain ID
raw.files <- get.pdb(ids, path=tmp.dir)

### Split PDB files by chain ID
files <- pdbsplit( raw.files, ids, path=tmp.dir)

... and the files are deleted when you leave R

[Barry Grant]

The tmpdir() is much better and indeed the 53 transducin guys is way to many for NMA.

Also the transducin example in the pca demo is great. Perhaps we can take just the four mentioned in the manuscript example from the transducin object for NMA? Not a big deal - I like the kinase example too so lets leave it for now...

[Barry Grant]

I think it would be best if all demos did not require external programs to be setup in order to run - e.g. muscle for alignment and dssp for sse. I think both the nma and pca ones need muscle currently.

One solution is to use the transducin data (or a subset there of) as mentioned above. Otherwise we need a way to tell a novice user whats wrong if the demo fails at this point.

BTW. Are demos tested by CRAN? If so this becomes a bigger issue.

[Lars Skjærven]

It doesn't seem that the demo directory is tested. it's certainly not by R CMD check.. and I think cran just uses "check".

Hmm.. Ok. I can change this, but I feel that it's not so much of a demo if you load half of the data in. Doing nma.pdbs(transducing$pdbs) is certainly not an option, and filtering out just a few structures would include some messing around with the pdbs object which can also be confusing for a new user.

How about I make some better error reporting in functions that use muscle and dssp? hmm

[Barry Grant]

Good idea - better error checks for muscle early on would solve all this.

btw. I have been looking into Rstudio for teaching and the demo fails their but not in the terminal. The error is "sh: muscle: command not found"

Do others see this problem with seqaln, dssp etc. in Rstudio?

[Xinqiu Yao]

Never happened to me. Are you using Mac? I guess it is the shell called was not the one you normally use...

[Lars Skjærven]

error will now sound:

> dssp(pdb)
Error in dssp(pdb) : Launching external program 'dssp' failed
  make sure 'dssp' is in your search path

[Lars Skjærven]

I checked dssp() and pdbaln(). both seems to work in my rstudio, also on my mac

[Lars Skjærven]

should be OK with the warnings in the demos and error checking in dssp() and seqaln()