Similarity searching using large-scaled transcriptome data
has recently been introduced into chemical similarity comparison.However,
they mainly emphasized on global expression comparison.
It is inaccessible for researchers to see more deeply into underlying biology.
For example, what contribute to their similarities or in what biological
mechanisms do they resemble. Besides, investigation of side effort for
drugs is challenging in biomedicine, because in most cases the action
of a chemical is complicated. Most chemicals have multi-targets that
target to more than one pathway. It's hard for global comparison to
figure out the situation that two chemicals though do not act exactly
alike, share some part of mechanisms.
We devise a new method that makes some
improvements in considering above two points. We set foot on a modularized
concept and doing similarity comparison within small biological meaningful categories of genes.
That is partitioning genes into gene ontology categories and implement
similarity comparison within each category.
Based on our method a web-based service was developed for facilitating
researchers, especially experimenters to use. By far,
whole data from connectivity map(build 1) are used as basic library for
validating the rationality of our method and querying target chemicals. Persistent
efforts will be made in adding data from other sources, including different
disease state, other cell lines and organisms to upgrade , enlarge our data
resource and expand the system's applications. Result is given in textual
and graphical mode both supporting mail return.