Structure of Chemical Space and the Principle of Strong Causality

It is not sufficient to generate offspring around a parent structure in a systematic way. To enable a systematic search in chemical space, both the ordering of chemical space and the corresponding fitness landscape must be strongly related. This means that small steps in chemical space must be correlated with small changes in predicted fitness or experimentally determined bioactivity (Principle of Strong Causality 34 ). Otherwise the whole concept of optimization by evolutionary search will be...

Xr2

PCA eigenvalues A, --(5.6) 2> ,xf, ,4 Kurtosis - x L- where tim are elements of the score matrix after PCA, w, is the weight vector, and n is the number of atoms in the molecule. The above parameters are defined as directional descriptors (in a more recent version the third order moment, skewness, was replaced by an information content index 56 ). The use of PCA ensures they are invariant to the reference system. A set of non-directional descriptors, which are derived from the directional...

Alignment Free 3D Descriptors

Considerable efforts continue to be dedicated to the development of descriptors which, whilst still retaining 3-D information, are not dependent on how molecules are superimposed to each other. In this Section we focus our attention mainly on molecular moments (WHIM and CoMMA), descriptors derived from molecular surfaces (MS-WHIM and MS-ACOR) and descriptors derived from spectra (EVA and CoSA). The moments of atomic properties are used as descriptors by the WHIM 54 and CoMMA (Comparative...

Info

This simplified example illustrates the concept of mean centred activity in SOMFA. The example is set up such that a hydroxyl group enhances the activity while steric bulk adjacent to the hydroxyl reduces binding. Activities are mean centred so that high active compounds have positive values and low active compounds negative ones. Molecules are then superimposed using the common core and embedded in a grid. The mean centred activity of each molecule is finally added to every grid...

Qsar

For many years, 3-D QSAR has been synonymous with CoMFA (Comparative Molecular Field Analysis) 26 . CoMFA was the first technique to implement in a QSAR approach the concept that a specific biological activity of a molecule is an inherent property of its three-di mensional structure and that any binding between a receptor and a ligand is mostly the product of non-covalent weak interactions. The idea behind CoMFA is that the 3-D steric and electrostatic properties of a molecule can be described...

Topological Descriptors

In the past few years, there has been a consistent return to the use and development of 2-D approaches. This tendency was mainly driven by technologies like combinatorial chemistry and high-throughput screening, which typically make available large amounts of data, and require, more than anything else, easily computable descriptors. This Section will cover some of the most applied 2-D descriptors, in particular fragment-based descriptors, atom pairs, topological torsions, Kier and Hall...

The Targets used for Virtual Screening

Three different targets have been used in our virtual screening assays. According to the given knowledge base, the previously collected experience, and the reported structural properties of the three targets, different search strategies have been applied. 10.8.1 First Leads for tRNA-Guanin-TVansglycosylase by Searches with LUDI The first target, tRNA-guanine transglycosylase (TGT) catalyzes the initial step of the post-transcriptional modification in the anticodon loop of cognate tRNA,...

The Structure of the Search Space

The nature of the search space, the space of accessible chemical entities and their biological, physico-chemical and pharmacological properties has a large influence on the optimal choice of appropriate selection methods. Structure-property relationships (SPRs) of various kinds have been considered in Chapter 8. Obviously one would use different selection methods if dealing with a unimodal (Chapter 8, Figure 8.4a) or multimodal (Chapter 8, Figure 8.4b) SAR. But what do real SARs look like...

Scoring Functions for Receptor Ligand Interactions

A very broad range of methods exists to estimate how strongly a molecule in a given pose will bind to a macromolecular target. Apart from simple measures of steric and electrostatic complementarity that are often used in the initial stages of docking or de novo design, most scoring functions estimate the free energy of binding of a receptor-ligand complex in aqueous solution. Only few methods, however, address the full thermodynamic cycle 52 involved in the binding process (see Chapter 10)....

Prediction of Physicochemical Properties

Physical properties are becoming recognized as important factors, which govern the ability of drugs to show an effect in vivo. There is little point in making very potent compounds which bind tightly to the receptor, if they cannot show their effect due to poor solubility, or poor bioavailability. Alongside this has come the realization that if the physical properties are not optimal, it is often difficult and time consuming to fix them later. Therefore, it is far better to start with something...

Fragment Based Similarity Searching

The first two reports of similarity searching appeared in the mid-1980s, based on work carried out at Lederle Laboratories 10 and at Pfizer 11 , The starting points for these two, near-contemporaneous studies were very different, but both groups of workers realized that counts of the numbers of fragment substructures common to a pair of molecules provided a computationally efficient, and surprisingly effective, basis for quantifying the degree of structural resemblance between the two molecules...

V

Examples of 3-D fragment descriptors. uses the distances between triplets of atoms. For each triplet in a molecule, the three interatomic distances are squared and summed to give a single value, and the distribution of these values is used to describe the molecule. Closely related to this is the atom triplet descriptor of Nilakantan et al. 57 . Here, the three distances between the three atoms comprising the triplet are sorted into increasing length the first is left alone, the...

Summary and Conclusions

In this Chapter we describe different schemes for performing a virtual screening study. A key point is a consecutive hierarchical filtering strategy to decompose a large database into a small sample of prospective hits ready for testing and co-crystallization with the target protein. Three successful case studies, using tRNA-guanine transglycosylase, human carbonic anhydrase, and aldose reductase are described in this paper, to suggest database entries as new lead structures. In a...

References

Taylor, in Novel Drug Delivery and its Therapeutic Implications, Prestcott, Nimmo (Eds.), John Wiley, Chichester 1989, pp. 33 4. 2 W. M. Pardridge, D. Triguero, J. Yang, P. Cancilla, J. Pharmacol. Exp. Ther. 1990, 253. 884-891. 3 D. A. Smith, in Computer Assisted Lead Finding and Optimisation, Wiley-VCH, 1997, pp. 265-277. 4 D. E. Leahy, J. J. Morris, P.J. Taylor, A.R. Wait, J. Chem. Soc. Perk. Trans. 2 1992, 723-731. 5 C. Hansch, A. Leo, in Exploring QSAR...

Hansch Analysis

The classical QSAR approach (also known as Hansch analysis) correlates biological activity values with physical and chemical properties by linear, multiple linear or non-linear regression analysis (see Section 5.8.2.1). The first equation explaining a biological activity through the linear combination of parameters describing hydrophobic (logP) and electronic (a) effects was proposed by Hansch and Fujita in 1964 1 Logl C - k x logP + k2x.a + k (5.1) where C is the concentration that produces a...

Conclusions

This Chapter highlights the ability of pharmacophores to divorce the 3-D structural requirements for biological activity from the 2-D chemical makeup of a ligand. The resultant descriptors are thus able to exploit even limited data regarding a target to discover novel active chemotypes. The methodology can be alignment-independent if necessary, with calculation speeds rapid enough to permit their use on datasets deemed too large for most other 3-D descriptors. It is this combination of...

A

Full Library 6,681 4,288 28.6 M Figure 2.5. A REOS analysis of combinatorial library. Figure 2.5 shows the number of reagents that could be used in the synthesis of an aryl amide library from benzoic acids and anilines. There are more than 28 million possible combinations for this two-component library. The three filtering steps described above reduce the size of the library almost 30-fold. 2.4 Chemistry Space Methods The idea that similar compounds will have similar properties is one of the...