I am running foldx position scan for GFPmut2 (PDB: 6GO9) and I use a config file with the following template:

pdb-dir= <pdb location>
output-dir= <output directory>
output-file= <file name>
positions= <list of positions, comma separated: QA204R,KA107R...>

I had created three different config files with different position list.

Some of the positions were present in more than one files. Foldx produced different estimates of ΔΔG for these positions from these different config files. Example:

File1-LYSA41K   0
File1:LYSA41R   -0.353758
File2-LYSA41K   0
File2:LYSA41R   -0.0356187
File3-LYSA41K   0
File3:LYSA41R   -0.0031362

Another example

File1-LYSA52K   0
File1:LYSA52R   0.0057333
File2-LYSA52K   0
File2:LYSA52R   -0.00809613
File3-LYSA52K   0
File3:LYSA52R   0.0077189

I can certainly run foldx with a union of the three list of positions but I don't understand how this difference can even occur. Shouldn't each position scan be independent?


1 Answer 1


Minute values. So just a reminder that the ∆∆G units is a kcal/mol. Water bounces around and collides with different energies following a Maxwell-Boltzmann distribution, where the average collision at 25°C is 0.59 kcal/mol (thermal energy constant). A hydrogen bond reduced the ∆G by 1–3 kcal/mol. Other interactions such as π-π, π-S are on the 1–2 mark. So the difference you are seeing is totally negligible.

Heuristic. The reason why you see the differences is that the force field calculations are heuristic and are not testing all perturbations —it is not only sampling a limited amount of conformers of side chains (kind of like PyMOL would be if it were FF based), but neighbour perturbations. So sometimes you hit a wee energy minima. Hence why in some protocols replicates are required —not mutation scanning mind you.

Bonus: Protein pI. Lastly, a caveat about pI of protein, you probably already know this, but best be safe. Your scores are all low, so I am guessing these surface mutations? Improving thermal stability is great for increasing the T_M measured by DSC or catalysis if you are scanning a transition state as done for theozyme design, but it does not mean you get better solubility. In this case, you are introducing positive charges, which will alter the pI of the protein upwards and most protein have a pI less than 7 and the closer it is to pH 7 the more they crash out. If you are designing a high pI GFP, Rosetta supercharge does FF calculations to alter the pI of protein, which might be worth a gander.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.