6
R. Kusko and H. Hong
reliable knowledge than that provided by any individual data source for risk assessment of chemicals attracts attention of computational toxicologists [21]. Network
analysis-based algorithms have been developed for analyzing such large, diverse,
and sparse data in computational toxicology [22, 23]. To shed insight into this new
method, Chap. 5 presented a network-based systems pharmacology approach that
integrates the networks of proteins, genes, drug target, and the human protein–protein interactome for assessing the risk of drug-induced cardiotoxicity in humans.
MoA is the functional or anatomical change caused by chemicals, at the cellular
level or at the molecular level that is often used as mechanism of action [24]. It is
important knowledge for understanding toxicology of chemicals when the molecular target of chemicals has not yet been determined. It can be used to guide development of predictive models in computational toxicology. Chapter 6 introduced a
MoA-guided novel computational toxicology approach that is based on molecular
modeling and is implemented in the target-specific toxicity knowledgebase (TsTKb)
that contains a pre-categorized database of MoA for chemicals and provides pre-built
and category-specific predictive models.
Predictive models in computational toxicology are often developed based on many
molecular descriptors using different machine algorithms [25]. One of the key steps in
development is to select important descriptors. Chapter 7 discussed different methods for removal of redundant and irrelevant molecular descriptors to improve the
performance and interpretability of the model. The strengths and shortcomings of
some feature selection and extraction methods in current computational toxicology
practices were summarized.
Genomics is the study of genomes, including all molecules such as DNA and RNA
and their structures and functions. Adverse effect of a chemical could be caused by
the interactions between the chemical and the target genome such as human genome,
such is the scope of toxicogenomics [26]. Toxicogenomics has been widely applied
in current toxicology practices. A database spanning disciplines of toxicogenomics
is the DrugMatrix, which includes gene expression of some 600 therapeutics at
multiple doses and 96 signatures relating to phenotypes. Chapter 8 gave a comprehensive description of a legacy resource of toxicogenomics, DrugMatrix and its
automated toxicogenomics reporting system, the largest molecular toxicology reference database and informatics systems, which contains thousands of gene expression
datasets generated using different microarray platforms.
Given the increasing prevalence of toxicogenomics resources such as the DrugMatrix database, a methodology known as pair ranking was developed to compare
transferability between the systems used for testing. Chapter 9 introduced the pair
ranking (PRank) method that is developed for quantitative evaluation of assay transferability between the different toxicogenomics platforms.
Several computational toxicology approaches have emerged as a hybrid with computational chemistry. For example, molecular dynamics (MD) simulation was originally used in chemistry to detail interactions between chemicals and biological
molecules (including DNA and proteins). For computational toxicologists, MD simulation allows for surveillance of potential fluctuations or conformational changes
that a chemical might induce on a biomolecule [27]. Chapter 10 reviewed available
R. Kusko and H. Hong
reliable knowledge than that provided by any individual data source for risk assessment of chemicals attracts attention of computational toxicologists [21]. Network
analysis-based algorithms have been developed for analyzing such large, diverse,
and sparse data in computational toxicology [22, 23]. To shed insight into this new
method, Chap. 5 presented a network-based systems pharmacology approach that
integrates the networks of proteins, genes, drug target, and the human protein–protein interactome for assessing the risk of drug-induced cardiotoxicity in humans.
MoA is the functional or anatomical change caused by chemicals, at the cellular
level or at the molecular level that is often used as mechanism of action [24]. It is
important knowledge for understanding toxicology of chemicals when the molecular target of chemicals has not yet been determined. It can be used to guide development of predictive models in computational toxicology. Chapter 6 introduced a
MoA-guided novel computational toxicology approach that is based on molecular
modeling and is implemented in the target-specific toxicity knowledgebase (TsTKb)
that contains a pre-categorized database of MoA for chemicals and provides pre-built
and category-specific predictive models.
Predictive models in computational toxicology are often developed based on many
molecular descriptors using different machine algorithms [25]. One of the key steps in
development is to select important descriptors. Chapter 7 discussed different methods for removal of redundant and irrelevant molecular descriptors to improve the
performance and interpretability of the model. The strengths and shortcomings of
some feature selection and extraction methods in current computational toxicology
practices were summarized.
Genomics is the study of genomes, including all molecules such as DNA and RNA
and their structures and functions. Adverse effect of a chemical could be caused by
the interactions between the chemical and the target genome such as human genome,
such is the scope of toxicogenomics [26]. Toxicogenomics has been widely applied
in current toxicology practices. A database spanning disciplines of toxicogenomics
is the DrugMatrix, which includes gene expression of some 600 therapeutics at
multiple doses and 96 signatures relating to phenotypes. Chapter 8 gave a comprehensive description of a legacy resource of toxicogenomics, DrugMatrix and its
automated toxicogenomics reporting system, the largest molecular toxicology reference database and informatics systems, which contains thousands of gene expression
datasets generated using different microarray platforms.
Given the increasing prevalence of toxicogenomics resources such as the DrugMatrix database, a methodology known as pair ranking was developed to compare
transferability between the systems used for testing. Chapter 9 introduced the pair
ranking (PRank) method that is developed for quantitative evaluation of assay transferability between the different toxicogenomics platforms.
Several computational toxicology approaches have emerged as a hybrid with computational chemistry. For example, molecular dynamics (MD) simulation was originally used in chemistry to detail interactions between chemicals and biological
molecules (including DNA and proteins). For computational toxicologists, MD simulation allows for surveillance of potential fluctuations or conformational changes
that a chemical might induce on a biomolecule [27]. Chapter 10 reviewed available