Hi all,
I am trying to predict the structure of a peptide from a gene we've only just discovered (isolated to just this one fungal pathogen and found nowhere else). I've tried homology modelling with little to no success and we have NO structural or protein assay data. So I turned to I-TASSER.
My first prediction wasn't wonderful. With a C-score of -1.81, a TM-score of 0.50+-0.15 and an RMSD of 10.2+-4.6. I have seen a lot worse published i.e., C-score approx -2.8.
Having a lot of experience in molecular dynamics I decided to take the result from I-TASSER and convert it into a Gromacs format using the AMBER 99ffSB forcefield. I have an HPC sitting around doing very little so I took the structure and ran an MD simulation in water for a whole month. It got through about 900 ns of simulation time.
Taking the most visited/common/average protein structure from a converged (on RMSD) region of the simulation trajectory I decided to us this for the "Specify template without alignment" entry during a resubmission with the original sequence. The results have come back and they are an improvement on the first: C-score of -0.94, a TM-score of 0.60+-0.14 and an RMSD of 8.1+-4.4. With the following returned top PDB hits:
PDB TM Score RMSD IDEN COV
3m62A 0.549 5.53 0.089 0.861
4fgvA 0.492 5.25 0.079 0.754
3nmzA 0.487 4.93 0.064 0.715
I am not overly familiar with the inner workings with I-TASSER. Have I skewered my second prediction by feeding in a "guessed" geometry?
Thanks