Current Release of BLOCKS

BLOCKS Database Version 14.0, October 2003
Copyright 2003 by Fred Hutchinson Cancer Research Center
1100 Fairview AV N, A1-162, Seattle, WA 98109

Version 14.0 of the BLOCKS Database consists of 24,294 blocks representing 4944 groups documented in InterPro 6.0 keyed to SWISS-PROT 41.0 and TrEMBL 23.0 obtained from the InterPro server .

The BLOCKS Database is based on InterPro entries with sequences from SWISS_PROT and TrEMBL and with cross-references to PROSITE and/or PRINTS and/or SMART, and/or PFAM entries.

The BLOCKS Database was constructed by the PROTOMAT system (S Henikoff & JG Henikoff, "Automated assembly of protein blocks for database searching", NAR (1991) 19:6565-6572) using the MOTIF algorithm (HO Smith, et al, "Finding sequence motifs in groups of functionally related proteins", PNAS (1990) 87:826-830) as implemented in Block Maker.

To avoid using possible false positive sequences added to the InterPro entries automatically (without human oversight), BLOCKS were made for each InterPro entry using just the sequences in SWISS-PROT, and then TrEMBL sequences were added if they fit the resulting BLOCKS model.

InterPro 6.0 consisted of 7751 entries. The 4944 entries of these represented in BLOCKS 14.0 were selected as follows:

 7751
-1906 entries with fewer than 3 SWISS-PROT sequences eligible for PROTOMAT (1)
 -901 entries with no PROSITE, PRINTS, SMART or PFAM component (2)
 -156 entries participating in InterPro parent/child relationships (3)
 4788

+ 204 entries PRINTS-only entries with fewer than 3 SWISS-PROT sequences (4)
 4992


 1354 blocks entries taken from PRINTS (4)


 3638 entries processed by PROTOMAT
  - 4 entries with too many sequences to process with PROTOMAT
  - 6 entries for which PROTOMAT failed to find blocks
  -38 entries for which final blocks were obviously useless (5)

 3590 blocks entries made by PROTOMAT


NOTES:
(1) PROTOMAT requires at least 3 sequences to make blocks. To be more
confident that the sequences used are actually members of the InterPro
protein family, we used only sequences from SWISS-PROT. Then, to reduce
redundancy, we use only the longest SWISS-PROT sequence among those
with the same gene name (characters before the "_" in the SWISS-PROT ID)
and similar organism name (first three characters following the "_").
For example, if an InterPro group included SWISS-PROT sequences named
AANT_HDVAM|P25989     LENGTH=214
AANT_HDVD3|P29996     LENGTH=195   
AANT_HDVWO|P29997     LENGTH=205   
only AANT_HDVAM would be used by PROTOMAT.

(2) InterPro now contains entries from several other sources. However,
these four sources tend to define a protein family in terms most amenable
to the BLOCKS model which is short, highly conserved regions. In
particular, PROTOMAT will generally produce unsatisfactory results for
groups comprised of a few, long, globally alignable sequences.

(3) Several InterPro entries are arranged into parent/child hierarchies
where all the sequences in a child entry are included in the parent
entry. Since PROTOMAT will tend to find the same blocks for the parent
and children, each major branch of a hierarchy is represented by only
one BLOCKS entry.

(4) Because the PRINTS model is the same as the BLOCKS model and PRINTS is
a curated collection of alignments, the PRINTS blocks were used directly 
for InterPro entries with only a PRINTS component as long as the PRINTS blocks 
had at least three sequences from any source. Then additional sequences
were added from TrEMBL if they fit the PRINTS model.

(5) These entries tend to be sites (e.g. IPR000886, IPB001216),
repeats (e.g. IPR000479, IPR001473) and viral proteins (e.g. IPR000208,
IPR000752).

Please note: The PROSITE pattern is not used in any way to make the BLOCKS Database and BLOCKS made from an InterPro PROSITE group may or may not contain the PROSITE pattern. Similarly, the SMART and PFAM multiple alignments are not used in any way to make the BLOCKS Database and BLOCKS made from an InterPro PROSITE, SMART or PFAM group may or may not overlap with the multiple alignments in those databases.


BLOCKS without compositionally biased blocks

To avoid the over-representation of compositionally biased blocks in search results, this subset of the BLOCKS Database excludes several biased blocks. It may give better results, especially with DNA queries.
BLOCKS home
Contact us

Page last modified October 2003