Any idea why there is a difference?

[Petr Škoda]

The p2rank provide different predictions for those two PDB files:

structure-full.pdb.txt
structure-chains.pdb.txt

The chain file was created only by selecting atoms on chains A, B (all chains in the full PDB)
Used command:

'/opt/p2rank-runtime/p2rank.sh predict -c /opt/p2rank/p2rank_2.1/config/default -threads 1 -f /data/p2rank/task/database/v2/19HC/working/structure.pdb -o /data/p2rank/task/database/v2/19HC/working/p2rank-output --log_to_console 1

Executed with https://github.com/rdk/p2rank/releases/download/2.1/p2rank_2.1.tar.gz

Any idea why this can be? Does p2rank utilize non-polymer atoms in any way?