Your title says holistic. This is a tad problematic as there's layers upon layers. Say, post-translation regulation, inhibiting metabolites, interacting protein etc. That is why when talking of an enzyme inhibitor, at the biochemical level one speaks in terms of k_i (inhibition constant), while at the cellular level ("holistic") one talks of IC50.
This could be a biocatalysis/synthetic biology problem (make an enzyme better) or translational medicine problem (why is this mutation pathogenic).
In terms of in silico enzyme design, one can have all sorts of issues that need to be addressed:
- catalysis
- side reactions
- stability
- aggregation
- solubility
Also there's the expression problems, which may be due to subpar promoter, insufficient cofactors and poor codon usage.
For clinical variant assessment, a variant may:
- destabilise leaving to degradation
- remove a post-translational modification
- disrupt the active site, affecting catalysis
- lose a binding partner
- result in incorrect localisation
- result in solubility issues
So force-field scores are not the sole solution. Furthermore, a forcefield method gives one a ∆∆G (difference in relative Gibbs free energy), not a K_M (Michaelis constant) and a k_cat (catalytic rate), although it is somewhat possible to calculate these, or a relative variant of these.
a variant with just 1 or 2 variants
A variant protein can have one or two mutations (e.g. G10R) in its sequence. A mutant and a variant are near synonyms, but variant is generally intended to be a mutant in a population.
Does these application be used for such purpose?
For enzyme design MD simulations are not generally the first step. Rosetta pmut_scan for example is a nice starting step as the mutational landscape of a protein is N^19, while a MD simulation takes weeks and 6+ core per single variant.
This calculates only an increase in stability (relative difference in Gibbs folding energy), not reactivity. For that you can use a transition state, whose results are inaccurate, but majorly affected by the set-up, i.e. junk-in junk-out plays a big part. There's a saying 'a week in the lab saves you an hour at the library': this is backwards for sarcasm as it happens too often, namely that is background research is always a lifesaver. An QM-MD simulation on the wild type reaction definitely can help in elucidating things if it is not clear.
Online tools
Generally, however, there are resources out there to do many stability calculations already.
SynBio If you are interested in picking a residue or two to make an enzyme more stable, without resorting to a thermophilic enzyme, PROSS is a very nice tool.
Clinical (Caveat: If your variant is a variant that affects you, a friend or a relative, you/they should speak to a genetic councillor and not do research on it yourself) There are several tools that can describe a variant at the protein level. Do remember that ∆∆G is a potential so more negative is more stable —except in SDM where the score is inverted.
