Table 1.

The databases used by the FFAS03 server

Database	Source of data	Preparation
NR85S (sequences)	NCBI, SEED	Protein sequences from the NCBI NR database and predicted open reading frames from unfinished bacterial genomes (kindly provided by Ross Overbeek) are clustered at 85% of sequence identity with the CD-HIT program (15). Regions of low complexity are masked with SEG (16).
PDB (profiles)	Protein Data Bank	FFAS profiles of all unique proteins (clustered at 99% identity level) from the PDB (17), including prereleased entries.
PFAM (profiles)	PFAM website	FFAS profiles of all PFAM (18) domains longer than 25 residues.
COG (profiles)	NCBI	FFAS profiles of all domains from COG database longer than 25 residues (19).
SCOP (profiles)	SCOP–ASTRAL website	FFAS profiles of SCOP domain sequences with <40% sequence identity to each other. SCOP protein sequences clustered at 40% of sequence identity have been downloaded from the Astral website (20).
JCSG (profiles)	JCSG website	FFAS profiles of all sequences of active targets of the Joint Center for Structural Genomics (21).