Hi, I was trying both Rosetta3.1 and Rosetta3.2 to do clustering. I had normal result from Rosetta3.1. But Rosetta3.2 failed to work. I used the same data set (generated by Rosetta2.3.0 abrelax mode), option and command, and the path is correct. I also got the error messege:
ERROR: Illegal attempt to score with non-identical atom set between pose and etable
ERROR:: Exit from: src/core/scoring/etable/EtableEnergy.cc line: 80
And in the log file, it says:
core.io.pdb.file_data: [ WARNING ] discarding 13 atoms at position 1 in file S_gr10_102.pdb. Best match rsd_type: LEU_p:NtermProteinFull
core.io.pdb.file_data: [ WARNING ] discarding 2 atoms at position 2 in file S_gr10_102.pdb. Best match rsd_type: GLY
core.io.pdb.file_data: [ WARNING ] discarding 2 atoms at position 3 in file S_gr10_102.pdb. Best match rsd_type: GLY
(basically every residue)
My option file is like this:
-database /...(my working folder)/rosetta-3.2/rosetta_database/
What could be the problem? Does Rosetta3.2 require some kind of special idealization of the input structure before clustering?
I'm not aware of any changes from 3.1 to 3.2 in clustering - maybe just use 3.1 if it works.
The atom discarding warnings, and the atom type set mismatch error, are indicative of a swap between centroid mode and fullatom mode that you did not intend. I assume your inputs are fullatom but it's trying to run in centroid. Try the in:file:fullatom flag?
The in:file:fullatom flag works. Thanks!
Now I know that Rosetta3.2 cluster application accept centroid-mode structure file by default. Is it also true for other Rosetta3.2 applications?
Many of the abinitio/relax related primary and secondary applications have assumptions of centroid inputs.
Docking and loop modeling have an early centroid phase and are compatible with either type of input.
Other applications generally assume full atom inputs.
There is no consistency (and of course no documentation) about this, unfortunately...