Mining Spatial Association Rules for Composite Motif Discovery - Mathematical Approaches to Polymer Sequence Analysis and Related Problems - page 104

edited by the user through a graphical user interface. This wrapper is also in charge
of automatically setting the min parameter as per Algorithm 1 in Sect. 5.4.1 .
By merging consecutive bins through the rules in D I , many spatial association
rules are discovered, which differ only in some intervals of inter-motif distances.
An unmanageably large number of association rules makes interpretation of results
cumbersome for the biologist. For this reason, association rules are filtered before
being shown to the user. Three filtering criteria are considered. The first criterion
selects the association rules with the smallest bins among rules with the same motifs,
the same confidence and supported by the same sequences. The second criterion
selects the association rules with the greatest support among those with the same
motifs and confidence, whose bins are included in the bins of the selected rules and
whose list of supporting sequences is included in the list of supporting sequences
of the selected rules. The last criterion selects the association rules with highest
confidence among those with the same motifs, whose bins are included in the bins
of the selected rules and whose list of supporting sequences is included in the list of
supporting sequences of the selected rules.

5.5 Case Study

To show the potential of the integrated system, a pilot study is conducted on trans-
lation regulatory motifs located in the nucleotide sequences of untranslated regions
(UTRs) of nuclear transcripts (mRNAs) targeting mitochondria. These motifs are es-
sential for mRNA subcellular localization, stability and translation efficiency [ 50 ].
Evidence from recent studies supports the idea that the nature and distribution of
these translation regulatory motifs may play an important role in the differential
expression of mRNAs [ 11 ].
Datasets are generated as a view on three public biological databases, namely
MitoRes, 6 UTRef and UTRsite. 7 The view integrates data on UTR sequences and
their contained motifs, together with information on the motifs width and their start-
ing and ending position along the UTR sequences in the UTRminer [ 48 ] database.
We base our analysis on a set T of 728 3 0 UTR sequences relative to the human
species. Twelve motifs are initially considered (set M). By setting min D 4, several
frequent motif sets (set S ) are extracted in the first phase. We focus our attention on

with the largest support set (111 3 0 UTR sequences). It contains
three motifs, which are denoted as x, y and z . The hierarchy defined on motifs has
three levels (Fig. 5.2 ), but we consider only the middle level, since the top level con-
veys little information on the constituent motifs of a module, while the bottom level
is too specific to find interesting rules.

the motif set S 2 S

6 http://www2.ba.itb.cnr.it/MitoRes/
7 http://utrdb.ba.itb.cnr.it/