352
A. Zakharov and A. Lagunin
Biological pathways and biological processes can characterize mechanisms of
drug side effects at more general level than the target molecules. Their assessment
is based on prediction of drug interactions with the targets and analysis of druginduced changes in gene expression profiles. In the first case, assessment of profiles
in drug interactions with proteins is made for a set of compounds, some of which
cause, and others do not cause side effects. Each of the predicted protein is associated with the biological pathway (signaling, metabolic, and regulatory). Every
biological pathway receives a score, which is calculated as the sum of probable
interactions with proteins that are a part of the way for all compounds causing an
appropriate side effect. The same estimation is calculated for the compounds that
do not cause this side effect. After that the ratio of estimates is calculated. Pathways
are considered associated with a side effect if the ratio is greater than 1, or the way,
the interaction with which is predicted only for compounds that cause the side effect [63].
Drug-induced changes in gene expression profiles can be used for searching of
the biological processes associated with side effects by the gene set enrichment
analysis [64]. In this method a set of genes involving in a biological process is created. Then, a list of estimates for changes in the gene expression is calculated based
on comparison of gene expression after the compound action and in the norm. The
list is sorted by ascending or descending the estimates subject to direction in which
the gene expression is changed (hyper-or hypo-expression). The basic analysis hypothesis is that the genes involved in the same biological process should be clustered mainly on the top or bottom of the sorted list.
In 2006, Lamb and co-authors introduced Connectivity Map (CMap) as a phenotypic-based drug discovery approach based on comparison of the disease gene
signature and drug-induced changes in gene expression profiles [65]. It was shown
that CMap approach can be used to reveal side effects of drugs [66]. The constructed multigene expression signature was used to predict future onset of the proximal
tubular injury in rats [67]. CMap approach has a limitation due to its applicability
only for drugs having experimentally determined drug-induced changes of gene
expression and it cannot be used for new drugs or new drug-candidates. This limitation may be partly overcome by prediction of possible drug-induced changes in the
gene expression for new drug-like compounds on the basis of existed experimental
microarray data. Such possibility is realized on a freely available DIGEP-Pred webservice (http://www.way2drug.com/GE). It also provides the links between gene
names in the predicted drug-induced changes in the gene expression and Comparative Toxicogenomics Database [68] which simplifies interpretation of predicted
results due to the access to relationships of genes with diseases, side effects and
biological pathways [69].
One of most important parameters considered during the drug development is
assessment of a potential participation of drugs in drug-drug interactions. Usage of
multiple drugs is a common practice in the treatment of many diseases, including
cardiac insufficiency (diuretics and vasodilators), malignant neoplasms and some
infectious diseases (HIV, hepatitis C, etc.). Despite a positive effect of the simultaneous usage for several drugs, there is a risk of negative influence on each other
A. Zakharov and A. Lagunin
Biological pathways and biological processes can characterize mechanisms of
drug side effects at more general level than the target molecules. Their assessment
is based on prediction of drug interactions with the targets and analysis of druginduced changes in gene expression profiles. In the first case, assessment of profiles
in drug interactions with proteins is made for a set of compounds, some of which
cause, and others do not cause side effects. Each of the predicted protein is associated with the biological pathway (signaling, metabolic, and regulatory). Every
biological pathway receives a score, which is calculated as the sum of probable
interactions with proteins that are a part of the way for all compounds causing an
appropriate side effect. The same estimation is calculated for the compounds that
do not cause this side effect. After that the ratio of estimates is calculated. Pathways
are considered associated with a side effect if the ratio is greater than 1, or the way,
the interaction with which is predicted only for compounds that cause the side effect [63].
Drug-induced changes in gene expression profiles can be used for searching of
the biological processes associated with side effects by the gene set enrichment
analysis [64]. In this method a set of genes involving in a biological process is created. Then, a list of estimates for changes in the gene expression is calculated based
on comparison of gene expression after the compound action and in the norm. The
list is sorted by ascending or descending the estimates subject to direction in which
the gene expression is changed (hyper-or hypo-expression). The basic analysis hypothesis is that the genes involved in the same biological process should be clustered mainly on the top or bottom of the sorted list.
In 2006, Lamb and co-authors introduced Connectivity Map (CMap) as a phenotypic-based drug discovery approach based on comparison of the disease gene
signature and drug-induced changes in gene expression profiles [65]. It was shown
that CMap approach can be used to reveal side effects of drugs [66]. The constructed multigene expression signature was used to predict future onset of the proximal
tubular injury in rats [67]. CMap approach has a limitation due to its applicability
only for drugs having experimentally determined drug-induced changes of gene
expression and it cannot be used for new drugs or new drug-candidates. This limitation may be partly overcome by prediction of possible drug-induced changes in the
gene expression for new drug-like compounds on the basis of existed experimental
microarray data. Such possibility is realized on a freely available DIGEP-Pred webservice (http://www.way2drug.com/GE). It also provides the links between gene
names in the predicted drug-induced changes in the gene expression and Comparative Toxicogenomics Database [68] which simplifies interpretation of predicted
results due to the access to relationships of genes with diseases, side effects and
biological pathways [69].
One of most important parameters considered during the drug development is
assessment of a potential participation of drugs in drug-drug interactions. Usage of
multiple drugs is a common practice in the treatment of many diseases, including
cardiac insufficiency (diuretics and vasodilators), malignant neoplasms and some
infectious diseases (HIV, hepatitis C, etc.). Despite a positive effect of the simultaneous usage for several drugs, there is a risk of negative influence on each other