Elsevier

Drug Discovery Today

Volume 9, Issue 3, 1 February 2004, Pages 127-135
Drug Discovery Today

Review
Early prediction of drug metabolism and toxicity: systems biology approach and modeling

https://doi.org/10.1016/S1359-6446(03)02971-4Get rights and content

Abstract

Many of the drug candidates that fail in clinical trials are withdrawn because of unforeseen effects of human metabolism, such as toxicity and unfavorable pharmacokinetic profiles. Early pre-clinical elimination of such compounds is important but not yet possible. An ideal system would enable researchers to make a confident elimination decision based purely on the structure of a new compound, and incorporate and use multiple pre-clinical experimental data to support such a decision. Currently available resources can be split into three categories: (i) structure–activity relationships (SAR) computational models based on compound structure; (ii) ‘pattern’ databases of tissue or organ response to drugs, compiled from high-throughput experiments; and (iii) ‘systems biology’ databases of metabolic pathways, genes and regulatory networks. In this review, we outline the advantages and drawbacks of each of these systems and suggest directions for their integration.

Section snippets

In-silico prediction of drug metabolism and related toxicity – the current state of affairs

A major step in assessing the potential toxicity of a novel chemical entity is the prediction of entry and fate of the compound in human metabolism. There are two distinct sides to this problem. The first task is to predict nominal metabolic transformations of a molecule. The second is to understand which enzymes will actually be involved, predict relative importance of alternate reaction routes and finally, their interactions with endogenous metabolism.

Over the years several computational

QSAR modeling

Much of the work in QSAR modeling has been focused on potential interactions of small molecules with major classes of proteins such as drug metabolizing enzymes 10, 11, 12, transporters [13], channels [14] and receptors 15, 16. These models usually correlate physicochemical descriptors with measured properties such as compound–enzyme binding and substrate potential. In a series of recent reviews and publications, QSAR models for substrates of several major drug-metabolizing cytochromes have

High-throughput ‘new biology’ and toxicology

During the 1990s, a new set of methods were developed to greatly increase the amount of experimental data from one experiment. These methods include automated DNA sequencing, microarray analysis of gene expression and protein profiling. Although these techniques have a wide variety of applications, we will consider only those related to toxicology here.

System-level reconstruction of endogenous pathways

Systems biology was defined by its founders as the integration of experimental data generated by high-throughput platforms (genetic, transcriptomic, proteomic, metabonomic) to understand function through different levels of biomolecular organization [46]. In the narrowest sense, this term describes the computational systems and models that deal with the analysis of the vast experimental data accumulated during the past ten years of ‘new biology’. These systems have multiple applications in drug

Pathway databases

A handful of public and commercial efforts have systematically emphasized various aspects of general biochemistry, metabolism and cell signaling. The first tier of these projects is essentially reference databases. These resources provide access to data on individual enzymatic reactions and their parameters (BRENDA [47], EMP [48]), multi-step metabolic pathways (MPW [49], EcoCyc and HumanCyc [50]) or human-curated maps for entire functional blocks [KEGG [43] and Biocarta (http://www.biocarta.com

Pathway reconstruction platforms

This second class of resource provides an added functionality to pathway databases by reconstructing condition-specific pictures of relevant pathways. These reconstructions, dubbed metabolic reconstructions (MR), have been pioneered by Selkov and co-workers 50, 51. Metabolic blocks corresponding to the genetic component of the organism are connected into wire diagrams for each different bacterial species. For some better-studied bacterial species, such as Escherichia coli and Bacillus subtilis,

Next steps – predictive ADME/Tox in the systems biology era

The theoretical and experimental approaches described in this review address different aspects of assessing metabolism and toxicity of novel drug candidates. What is needed is a platform that integrates different types of in silico modeling with high-throughput molecular assays and that enables the reconstruction of a system-level picture of the behavior of a drug. We envision the next generation platforms for predictive ADME/Tox as integrated systems that can process different types of inputs

Concluding remarks

In summary, we believe that the data and methodologies exist to greatly improve ADME/Tox prediction of novel chemical compounds early in the drug discovery process. The integration of modeling, high-throughput and systems biology approaches will allow true breakthroughs in in silico ADME/Tox assessment.

References (59)

  • J.L. Griffin

    Metabolic profiling of genetic disorders: a multi-tissue NMR spectroscopic and pattern recognition study into dystrophic tissue

    Anal. Biochem.

    (2001)
  • L.L. Smith

    Key challenges for toxicologists in the 21st century

    Trends Pharmacol. Sci.

    (2001)
  • M.W. Covert

    Metabolic modeling of microbial strains in silico

    Trends Biochem. Sci.

    (2001)
  • J. Talafous

    META. 2. A dictionary model of mammalian xenobiotic metabolism

    J. Chem. Inf. Comput. Sci.

    (1994)
  • G. Klopman

    META. 1. A program for the evaluation of metabolic transformation of chemicals

    J. Chem. Inf. Comput. Sci.

    (1994)
  • Darvas, F. (1987) MetabolExpert, an expert system for predicting metabolism of substances, in QSAR. In Environmental...
  • N. Greene

    Knowledge-based expert systems for toxicity and metabolism prediction: DEREK,StAR and METEOR

    SAR QSAR Environ. Res.

    (1999)
  • P.N. Judson

    Using new reasoning technology in chemical information systems

    J. Chem. Inf. Comput. Sci.

    (1996)
  • G. Klopman

    META. 3. A genetic algorithm for metabolic transform priorities optimization

    J. Chem. Inf. Comput. Sci.

    (1997)
  • S. Hayashi

    Genetic polymorphisms in the 5(-flanking region change transcriptional regulation of the human cytochrome P450IIE1 gene

    J. Biochem. (Tokyo)

    (1991)
  • S. Kato

    Cytochrome P4502E1 (CYP2E1) genetic polymorphism in a case-control study of gastric cancer and liver disease

    Pharmacogenetics

    (1995)
  • D. Korolev

    Modeling of human cytochrome p450-mediated drug metabolism using unsupervised machine learning approach

    J. Med. Chem.

    (2003)
  • D.F. Lewis

    An improved and updated version of the compact procedure for the evaluation of P450-mediated chemical activation

    Drug Metab. Rev.

    (1998)
  • S. Ekins

    Three-dimensional quantitative structure-activity relationship for inhibition of human ether-a-go-go-related gene potassium channel

    J. Pharmacol. Exp. Ther.

    (2002)
  • S. Ekins

    A ligand-based approach to understanding selectivity of nuclear hormone receptors PXR, CAR, FXR, LXRalpha, and LXRbeta

    Pharm. Res.

    (2002)
  • S. Ekins et al.

    A pharmacophore for human pregnane X receptor ligands

    Drug Metab. Dispos.

    (2002)
  • S. Beken

    Effects of L-proline on phase I and phase II xenobiotic biotransformation capacities of rat and human hepatocytes in long-term collagen gel cultures

    Altern. Lab. Anim.

    (2001)
  • H. Jackson

    Baculovirus-mediated high level expression of a human thiopurine methyl transferase

    Biochem. Mol. Biol. Int.

    (1996)
  • R.S. Thomas

    Identification of toxicologically predictive gene sets using cDNA microarrays

    Mol. Pharmacol.

    (2001)
  • Cited by (0)

    View full text