- This topic has 2 replies, 3 voices, and was last updated 12 years, 6 months ago by Anonymous.
April 23, 2011 at 12:56 am #882Anonymous
I’m using the standalong version of Rosetta Antibody. I can rebuild the antibodys which are already in PDB, however, for my own sequences, I can not build them for some reasons. I checked the results of the CDR Graft procedure. I found the “FR02.pdb” was distorted. Apparently, the some atoms of H-CDR3 residues were far away from the framework. Besides, I got a “VERY LOW CONFIDENCE PREDICTION” warning on the H3 loop.
I know this is because of the low homology to other antibodies, however, to my knowledge, it should not give me a distorted structrure. Since I got a distorted FR02, I can only build a distorted antibody whose H chain was broken at Q105 (which is the position of the H3 should be grafted to).
For those antibodys whose structures were already resolved, I could build them perfectly with RosettaAntibody
April 28, 2011 at 10:21 pm #5499Anonymous
I am sure that RosettaAntibody can predict sequences for which structures do not exist in the PDB. It is difficult to imagine that the 1500 jobs that the server has executed are for sequences which already have a structure in the PDB! I suspect that there is an issue with the sequence that has been used as an input. Probably it does not have one of the key signatures responsible for detecting the start/stop of CDR regions -resulting in a distorted grafted model (FR02.pdb). I recommend that you test you sequence for antibody specific signatures by submitting them to the Abnum server: http://www.bioinf.org.uk/abs/abnum/
May 26, 2011 at 2:24 pm #5659Anonymous
Maybe you have launched only the preparatory script:
perl -P $ROSETTA_ANTIBODY_PATH/scripts/ram_buildloop_wrapper4.pl pdb1xyz_chothia.pdb bit ram ram 1 1 1 1 1 0 0 1 1 0 0 0 0 FR02 2 1 3 2000 2000 3 0 query_l.fasta query_h.fasta
after that you’ll have to move to the ./build directory and then launch the ./1xyz_build_loops.bash script that will start the actual modeling (by default it will produce 2000 models, in some days of calculations).
However i would like to point out that the scripts coming with Rosetta2.3 (antibody in particular) are full of funny local paths, you have to found, and replace.
I don’t uderstant why : to launch any script you need to have all the “path.txt” or “rampath.txt” stuff, and then some part of the script doesn’t use the path given by user, making the debugging quite tedious.
I’ve also found a problem with the precompiled version of blast coming with my distribution (Ubuntu 11.04):
blastall -p blastp versus the fragments database does give a “No Hit Found” even if (checking by hand) the target sequence was present!!
Resolved by recompiling ncbi-blast.
But it’s still not clear to me, why i have to download the full non-redundant blast database in order to model an antibody, given the presence of antibody sequences database.
However I’ve found that when the fragment alignment doesn’t find any match for H3 in the fragment database it exit with error… it isn’t possible to try something else?
- You must be logged in to reply to this topic.