You are here

Ab Initio folding with HEM ligand molecule

1 post / 0 new
Ab Initio folding with HEM ligand molecule

Hello everyone,

I'm using Ab initio method to predict the structure of a protein with a HEMO group.

I just followed a couple of different links ( about how to introduce the external parameters of the ligand by using the ones in Rosetta (HEM.fa.params and HEM.cen.params) and other ones I generated in case the error I was experiencing was promoted by those parameters. For that reason, I generated a new ones (centroid and full-atom) by using the script

I was using some constraints for the full atom and the centroid, assigning different weights to each case and apparently, everything worked.

However, when I visualize the model generated, it's a little bit shocking to see how the N-terminus, of all models were generated (more than 1000), are in contact all the time with the HEM group.

When you open the generated models, you can quickly visualize how the first atom of the protein (nitrogen atom) start at the same coordinates as the first atom of the HEM group (in this case, the iron ion), and then, it's like all the sampling starts using both entities to be occupying the sample place. I also tried to translate the HEM, re-parametrize to see if this could be a plausible solution, but no success, the problem was not solved. 

I tried everything I found and I could imagine, but apparently, nothing works and I would like to know if anyone of you experienced the same problem in the past, and if yes, how it was solved, because documentation is pretty scarce for the Abinitio module and the examples found are pretty basic

The commands were used in the execution are detailed below:

AbinitioRelax.static.linuxgccrelease -database PATH2ROSETTA_DB \
    -fasta seq.fasta \
    -frag3  frags.200.3mers -frag9  frags.200.9mers \
    -out:file:silent seq_silent.out -out:pdb -out:path results/  \
    -constant_seed -jran `echo $RANDOM` \
    -loops:extended -loops:build_initial -loops:remodel quick_ccd -loops:refine refine_ccd -loops:relax fastrelax \
    -random_grow_loops_by 4 \
    -select_best_loop_from 1 \
    -nstruct 10  \
    -abinitio:relax -relax:fast \
    -extra_res_fa   HEM.fa.params \
    -extra_res_cen HEM.cen.params \
    -cst_fa_file native.cst -cst_file  native_CEN.cst \
    -cst_weight 50 -cst_fa_weight 5 \
    -ex1 -ex2 -extrachi_cutoff 10  >  AbInitio.log


Other protocols were tested as well, like this one, but with identical solution in terms of generated models:

AbinitioRelax.static.linuxgccrelease -database PATH2ROSETTA_DATABASE \
-seed_offset `echo $RANDOM` \
-run:protocol broker \
-nstruct 10  \
-fasta seq.fasta \
-frag3 frags.200.3mers -frag9  frags.200.9mers \
-out:file:silent seq_silent.out -out:pdb -out:path results/ \
-relax:fast \
-relax:jump_move \
-rg_reweight 0.0001 \
-packing:ex1 \
-packing:ex2 \
-fold_and_dock:move_anchor_points \
-fold_and_dock:set_anchor_at_closest_point \
-fold_and_dock:rigid_body_cycles 100 \
-fold_and_dock:rigid_body_frequency 0.01 \
-fold_and_dock:slide_contact_frequency 0.01 \
-extra_res_fa   HEM.fa.params \
-extra_res_cen  HEM.cen.params \
-cst_fa_file native.cst -cst_file native_CEN.cst \
-run:reinitialize_mover_for_each_job >  AbInitio.log


Any help will be really welcome.


And here goes the sequence, where the ligand is denoted in brackets with the letter code Z as suggested in other posts where ligand is included in the fast sequence


The sequence:




Many thanks in advance for any help

bad-model.png339.44 KB
Post Situation: 
Fri, 2021-03-05 06:49