Function PDB_Prepare

This function can benefit a lot from refactoring:

• split sub-functions into separate functions: these can be useful on their own;
• reduce the complexity of the code;
• optimize the code and remove dependency on dplyr;
• correct an ugly bug;

I will address all these topics below. The refactored code is also available on GitHub:

• PDB_Prepare function:
  : https://github.com/discoleo/R/blob/master/Chemistry/Proteins.Structure.FiScore.R
• separate Helper functions:
  : https://github.com/discoleo/R/blob/master/Chemistry/Proteins.Structure.R

PDB_Prepare Main Function

The refactored main function is in file Proteins.Structure.FiScore.R (this is only a convenience name for my script files; the initial name can be retained).

• most of the code has been moved to external helper functions;
• the lower limit of aa can be explicitly set: default = 5;
• the last for-loop should run more efficiently and the dependence on dplyr has also been removed;

Features

Are extracted by the helper function features.pdb (see file Proteins.Structure.R). This function can be used on its own and should be exported by the package).

• the function also uses the helper functions: as.type.helix and as.type.sheet;
• the structure name is stored as a factor (for efficiency): therefore requires explicit as.character() when used in the main function;

Torsions & B-Factor

Are computed by separate functions (see file Proteins.Structure.R). These functions can be used on their own as well.

• string extraction: the vectorized version is used directly and should run far more efficiently, e.g.:
  df_resno = as.numeric(stringr::str_extract(rownames(pdb_df), "[0-9]{1,}"));

Ugly bug was also corrected:

• the torsions function now stores an attribute with the complete cases (as a logical vector):
  attr(pdb_df, "complete") = isComplete;
• the BFactor function uses explicitly this information to select only the complete cases;

Other

• read.pdb: is a minor helper function not actually used in the code;

The refactored code should be faster and more robust. The function names are provisional and may be changed or adapted to better suite various workflows.

Note:

• the refactored code has NOT been thoroughly tested!

Helper function MINMAX_normalisation_func

The range (min, max) is computed multiple times. It should be computed only once. One of the following conditions also seems redundant: (r[1] == 0) && (r[2] == 0)), as min == max.

Also, inserting a space between "#" and the comments greatly increases readability.

### Helper functions for the analysis

# MIN-MAX normalisation based on the input array
MINMAX_normalisation_func = function(array) {
	# input = numeric array;
	# returns normalised array values;
	
	# check for cases where all B-factor values are 0;
	r = range(array);
	if((r[2] - r[1] == 0) && (r[1] == 0) && (r[2] == 0)) { return (0); }
	
	return ((array - r[1]) / (r[2] - r[1]));
}