I have two setups for a cluster of rod shaped proteins as a surface with water on the outside (it is a fibril of collagen protein, periodically stacked in the x-y dimension and a long water box in the z dimension - the x-y dimension is fixed to ensure a fibril environment through its periodic image, there are only two explicit proteins). In an effort to calculate the energetic contribution of a single point amino acid mutation had I pulled one protein out of the cluster and then performed umbrella sampling.
I was very lucky that for my first system the pulled protein never titled over the box dimension (the length of the collagen protein is longer than the width of either x- or y- box length - I am pulling along z). But unfortunately, in the second system this happened, giving the appearance of a continuous cluster (rather than a monomer) even when I had pulled a protein out of the cluster completely. Without significantly increasing my system size, by adding more proteins in the cluster, and further water molecules to ensure that the x-y plane will never cause PBC artefacts, I wish to calculate the free energy difference of association between the two systems (wildtype vs. single-point mutant).
My mutant amino acid is a bespoke type, I completely parameterised it myself. Would TI (Thermodynamic Integration) be the tool to slowly mutate from one amino acid type to another? And if so, how is this done in Gromacs?
Many thanks
Anthony
I am thinking perhaps Discrete Thermodynamic Integration (DTI) would be the best technique.
How about a non-periodic system (a water sphere) & free energy perturbation (wild-type mutant) . . .
Thanks for the reply Martin, unfortunately due to the fact that I need to replicate a fibril environment I need to rely on periodic boundary conditions. However, I have found a very interesting paper which may help with TI:
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4365728/
PMX is indeed good for protein mutations. If you are interested to TI with ligands have look to http://www.ncbi.nlm.nih.gov/pubmed/26544598. With TI you will have for sure an high accuracy DDG calculations but very time consuming. If you are interested to roughly estimate the DDG[wt->mut] you can consider to exploit MM/PBSA approach too. Have look here:
http://www.sciencedirect.com/science/article/pii/S0010465514003154
Hi Wojciech,
Thanks for the response. I've started work with pmx combined with the Crooks Fluctuation Theorem for a different system to the one described above. Essential, I am in the early stages of performing an Alanine scan in transmembrane protein dimer.
Unfortunately, after much discussion with the first author in the PMX publication, I won't be able to use it for the originally described system as it involves bond breaking and formation. Gromacs would struggle with its constraints implementation if I tried. I had thought of putting a bond length constraint on the two amino acids which morph into a single glycosylated residue, but I would then face big issue with charge (positive lysine and positive histidine).
Long story short. PMX for the transmembrane investigation, and back to umbrella sampling for the glycated collagen investigation.
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Biomolecules
- Proteins