Review
Lead discovery using molecular docking

https://doi.org/10.1016/S1367-5931(02)00339-3Get rights and content

Abstract

As the structures of more and more proteins and nucleic acids become available, molecular docking is increasingly considered for lead discovery. Recent studies consider the hit-rate enhancement of docking screens and the accuracy of docking structure predictions. As more structures are determined experimentally, docking against homology-modeled targets also becomes possible for more proteins. With more docking studies being undertaken, the ‘drug-likeness’ and specificity of docking hits is also being examined.

Introduction

Given the atomic resolution structure of a macromolecule, such as an enzyme, it should be possible to find novel molecules that bind to it, modulating its activity. This is the premise behind all structure-based ligand design and discovery efforts. Here we consider one aspect of this field: structure-based ligand discovery, emphasizing the screening of compound libraries using molecular docking.

The promise of docking is that the structure of the target will provide a template for the discovery of novel ligands, dissimilar to those previously known. One begins with a database of compounds and the structure of a receptor of interest and asks, ‘Of the compounds in the database, which is most likely to bind to the receptor?’ (Fig. 1). This apparently simple question disguises an enormous problem and a clever, if sometimes unsuccessful, strategy. The problem is that of predicting absolute binding affinities of many disparate molecules. Whereas predicting relative binding affinities for related molecules is possible, although time consuming, we have no reason to expect that we can predict absolute binding affinities for so many unrelated compounds. If docking has had an impact, it is because of its strategy of using databases of available compounds. This makes failure cheap; one simply goes to the next compound in a hit list. Docking for discovery is a screening technique: both false positives and false negatives are tolerated as long as true positives are found at a sufficiently high rate to justify the effort. How high should this hit rate be?

Section snippets

Comparing docking and high-throughput screening

Because high-throughput screening (HTS) is the dominant technique for pharmaceutical lead discovery, what level of hit-rate enhancement would be sufficient to justify pursuing structure-based docking? Indeed, if HTS is available, why do docking at all (Fig. 1)? Two recent studies begin to consider this question.

Using both HTS and virtual screening, inhibitors were sought for the type 2 diabetes target protein tyrosine phosphatase 1B (PTP1B) [1]. In the HTS experiments, a 400 000 compound

Recent applications of docking

Docking has been used to discover novel ligands for well over 30 targets. Work in the past year has continued to focus on enzymes (Table 1). The inhibitors discovered were novel, having little similarity to the known ligands. Most initial leads had affinities in the low-micromolar range. The new sulfonamide inhibitors of carbonic anhydrase II [3] are the exception—they are much more potent, but not completely novel. There has been an efflorescence of new docking methods in the past several

Integration of docking with design and virtual libraries

Can docking hits be turned into leads through synthetic elaboration? In several studies 5•., 10., the affinity of hits was improved by 10- to 1000-fold (Table 1), often through ‘classical’ structure-based techniques (i.e. beginning with a lead and using the complexed structure and chemical intuition to improve it). More sophisticated efforts have tried to include synthetic accessibility in the virtual screening from the beginning. In an effort to design novel CDK4 inhibitors, the de novo design

Technical advances in docking algorithms

Molecular docking continues to witness the introduction of new algorithms and programs—these are much needed given the weaknesses in conformational sampling and scoring. Along with efforts to improve established docking programs, such as autodock, dock, ecepp/prodock, flexx, flog, gold, green, icm, ludi, pro_leads, qxp and slide (reviewed in [21] and [22]), new docking programs have been published in the past year, including the eudoc algorithm [23], seed [24•], seeds [4•] and mm [25].

Two

The use of homology models in docking

For many interesting targets, an experimental structure is unavailable. In principle, homology modeling can calculate a structure for use in drug discovery [42], thereby dramatically increasing the number of targets to which docking might be applied. How reliable are these models for docking? This question cannot be answered definitively, but some tentative points may be made.

Several groups have used homology models to design or discover ligands for target proteins. Indeed, of the 12 docking

Hit conformation and promiscuous inhibitors

As is well known to screeners and medicinal chemists, many HTS hits are promiscuous and non-‘drug-like’. This can also be true of docking hits. Whereas detailed testing to confirm a hit routinely occurs in HTS projects, docking hits are not always confirmed as carefully. In our own painful experience, this can lead to artifacts that are confusing and time consuming. Every effort should be made to remove what we have come to call ‘pathological’ inhibitors from docking hit lists.

We can consider

Conclusions

The recent explosion of protein structures, and the advent of the genome projects, has renewed interest in using structure-based docking for early-phase lead discovery. Database docking has made considerable progress in the past decade, but it remains a screening technique. As such, current docking programs will dissatisfy investigators interested in definitive predictions of new inhibitors, and predictions of geometries can still go wildly wrong. In favorable circumstances, docking screens can

Acknowledgements

Supported by GM59957 (to BKS). SLM is a Northwestern University Presidential Scholar, and is partly supported by T32-GM08152 (K Mayo, PI) and a PhRMA Foundation Medical Student Fellowship. We are grateful to David Lorber for many discussions.

References and recommended reading

Papers of particular interest, published within the annual period of review, have been highlighted as:

  • • of special interest

  • •• of outstanding interest

Now in press

The work referred to in the text as (R Powers, BK Shoichet, unpublished data) is now published. The work referred to in the text as (BW & BKS, unpublished data) is now in press

References (61)

  • A. Schafferhans et al.

    Docking ligands onto binding site representations derived from proteins built by homology modelling

    J Mol Biol

    (2001)
  • G.M. Rishton

    Reactive compounds and in vitro false positives in HTS

    Drug Discov Today

    (1997)
  • Y.P. Pang et al.

    Discovery of a new inhibitor lead of adenovirus proteinase: steps toward selective, irreversible inhibitors of cysteine proteinases

    FEBS Lett

    (2001)
  • D.M. Freymann et al.

    Efficient identification of inhibitors targeting the closed active site conformation of the HPRT from Trypanosoma cruzi

    Chem Biol

    (2000)
  • T.N. Doman et al.

    Molecular docking and high-throughput screening for novel inhibitors of protein tyrosine phosphatase-1B

    J Med Chem

    (2002)
  • S. Gruneberg et al.

    Subnanomolar inhibitors from computer screening: a model study using human carbonic anhydrase II

    Angew Chem Int Ed Engl

    (2001)
  • T. Honma et al.

    Structure-based generation of a new class of potent Cdk4 inhibitors: new de novo design strategy and library design

    J Med Chem

    (2001)
  • M. Schapira et al.

    In silico discovery of novel retinoic acid receptor agonist structures

    BMC Struct Biol

    (2001)
  • I.J. Enyedy et al.

    Discovery of small-molecule inhibitors of Bcl-2 through structure-based computer screening

    J Med Chem

    (2001)
  • Y. Iwata et al.

    Discovery of novel aldose reductase inhibitors using a protein structure-based approach: 3D-database search followed by design and synthesis

    J Med Chem

    (2001)
  • S. Makino et al.

    Discovery of a novel serine protease inhibitor utilizing a structure- based and experimental selection of fragments technique

    J Comput Aided Mol Des

    (2001)
  • J.W. Liebeschuetz et al.

    PRO_SELECT: combining structure-based drug design and array-based chemistry for rapid lead discovery. 2. The development of a series of highly potent and selective factor Xa inhibitors

    J Med Chem

    (2002)
  • H.J. Boehm et al.

    Novel inhibitors of DNA gyrase: 3D structure based biased needle screening, hit validation by biophysical methods, and 3D guided optimization. A promising alternative to random screening

    J Med Chem

    (2000)
  • B.A. Grzybowski et al.

    Combinatorial computational method gives new picomolar ligands for a known enzyme

    Proc Natl Acad Sci USA

    (2002)
  • Siani MA, Skillman AG, Carreras CW, Ashley G, Kuntz ID, Santi DV: Development and screening of a polyketide virtual...
  • D. Joseph-McCarthy et al.

    Use of MCSS to design small targeted libraries: application to picornavirus ligands

    J Am Chem Soc

    (2001)
  • Mason JS, Cheney DL: Library design and virtual screening using multiple 4-point pharmacophore fingerprints. Pac Symp...
  • R. Olender et al.

    A fast algorithm for searching for molecules containing a pharmacophore in very large virtual combinatorial libraries

    J Chem Inf Comput Sci

    (2001)
  • A.M. Aronov et al.

    Virtual screening of combinatorial libraries across a gene family: in search of inhibitors of Giardia lamblia guanine phosphoribosyltransferase

    Antimicrob Agents Chemother

    (2001)
  • M.L. Lamb et al.

    Design, docking, and evaluation of multiple libraries against multiple targets

    Proteins

    (2001)
  • Cited by (427)

    View all citing articles on Scopus
    View full text