2006/03/29

DB got corrupted (again!) during weekend shutdown. Information-content and GO similarity measures must be recalculated. Replacement of HD in servers. Consider migrate DB to several smaller DBs (reduce update time).
Implementing new measure of similarity: compounds in a reaction are divided in substrates and products, similarity of two reactions according to their compounds maps substrates to substates and products to products only

2006/03/16

new pathways already in web. Queries take now aprox. 30 secs for 8 organisms and 24 pathways.
updating info on several metabolism-related pathways (SQLite database over 200MB)

2006/03/15

2006/03/14

2006/03/13

alpha value was de-activated, some distances were negative: fixed
web server v0.53: new organisms and pathways added, functionality of "Select/Deselect all" buttons improved

2006/03/12

2006/03/11

maximum distance criterion is not supported in Bio::Tree::DistanceFactory, removed from web interface
web server v.051: neighbor-joining clustering method added
web server v0.5: KEGG taxonomy of organisms included

2006/03/08

2006/03/07

web server v0.4: included optimizations for calculation of reaction distances: 8 paths & 8 orgs in 5 secs.

2006/03/06

2006/03/04

2006/03/03

started populating the DB. Table OrganismReaction was necessary (objects we are measuring in ReactionDistance are not Reaction but OrganismReaction).

2006/03/02

a DB created through the SQLite shell seems to have some compatibility problems when accessed from Perl scripts using DBI.
problems overriding system's Perl old modules: consider for mirror @ UPC.

2006/03/01

current DB schema could lead to wrong retrieval of enzyme sets in a reaction. Assuming the set of enzymes for a given reaction is the same for a specific organism, this problem can be solved by including "organism_id" as a FK in table Reaction and removing the (now) useless table OrganismReaction (the assumption holds for current dataset, KEGG documentation does not specify anything).

2006/02/28

2006/02/27

adding tables CompoundReactionDistance and EnzymeReactionDistance to support pre-calculation of reaction distances.
re-writing update-enzymes.pl: now includes calculation of enzymes in each organism from pathway files, calculation of which enzyme distances should be computed, computation of enzyme distances, and update of enzymes/enzyme distances in the DB.

2006/02/24

adding tables Compound, CompoundDistance, Reaction and ReactionDistance: everything to be pre-calculated.

2006/02/23

web server v0.2: enzyme distances are now stored as a SQLite DB. Queries are processed in roughly half the time: 22 secs for 8 orgs, 8 paths. Enzymatic similarity of two reactions is now calculated in O(3n) the number of enzymes, before it was O(n*n). Compuation of similarity of set of reactions can equally be reduced if we pre-calculate similarity of reactions.

2006/02/22

first test with enzyme distances stored in SQLite DB: works, but extremely slow, something is wrong.
more problems in www fixed: environment lacks most tools and Perl version in /usr/bin is older than it should be.
"SQLite.so: symbol fdatasync not found" problem in www fixed. DBD::SQLite Makefile should be changed to include two extra flags (-posix4 and -lrt) in gcc options (LDDLFLAGS and LDFLAGS). This problem seems to be exclusive of Solaris machines (once again, bravo!).
installed SQLite and DBD::SQLite. All enzyme distances (and probably reaction distances as well) will be moved to a DB to be accessed from the CGI.

2006/02/21

faster processing of queries: only pathway files of selected organisms are loaded (query with 8 orgs & 8 paths: 45 secs).
semi-automatic synchronization with KEGG: pathway files obtained from KEGG through scripts. Enzyme files obtained from pathway files. Distances not yet calculated obtained from enzyme files.

2006/02/16