Notas

Esquema

Modeling Molecules and Biomolecules: Basic Principles and Drug Engineering

Drug discovery systems 1 “In vitro” – “in vivo”: It is considered that the traditional [synthesis à screening à drug] process requires the availability of 5.103 to 5.106 compounds, 12 to 24 years and expenditures of US$ 300 to US$ 500 million for getting a single drug.

Drug discovery systems 2  “In silico” – “in vitro” – “in vivo”: Data base management Molecular models… or mixed techniques.

Current procedures of computer aided drug design: data mining It is understood that a huge availability of protein, gene and “small” molecule structures in data bases, after a patient mining, could yield results on several active molecule candidates. This is a matter of “system’s assembling” or a “puzzle” with the desired models.

Current procedures of computer aided drug design: molecular modeling The “state of the art” tools of computational chemistry can also provide new abilities with predictive models, where the scientists can create new structures “in silico”, perform virtual screenings, select “could be” candidates and provide experimentalist with a reduced set of molecules to be tested.

A model… Is this a jungle? It is “only” a model.

A molecular model

Computational Chemistry 1 Computational chemistry is the “chemistry of the dry lab” because it pursues to build theoretical models of molecular objects in the computer. Many physical, chemical and biological phenomena associated to them can be represented not only for stable but non-stable and transient species.

Computational Chemistry 2 The main purpose is to support experimental science, and even substitute expensive assays with cheaper computational models

Computational Chemistry 3   Among the most important properties that can be calculated are: Molecular structure, electron density and partial charges Energy relationships and spectra    All of them are very related to the biological effects of molecules. Particularly, molecular structure and electron densities are key points.

Current procedures of computer aided drug modeling: host and guest systems Proteins, nucleic acids, oligosaccharides, glycosides, lipids, etc. are typical “host” molecules. Ligand or “guest” systems could also be that kind of macromolecules, although drug design usually prefers seeking simpler molecules.

Current procedures of computer aided drug modeling: approaches Docking aproaches account complementarities of the host and the ligand Ligand approaches ignore the host and departs from previously known effects of families of compounds

Docking approach 1: It is known or modeled the host structure (either from experimental, homology or molecular modeling data) and a search is performed for the best ligand.

Docking approach 2: Docking approach searches for: Steric complementarity Electron density complementarity, that comprises electrostatic, dispersive and bonding effects. Hydrophobicity is a feature derived from the electron density complementarity.

Docking approach 3: Research for steric complementarity is usually performed by pattern recognition techniques, similarity analyses, and even by computer graphic tools. Electron density complementarity requires calculations of the hypersurfaces.

Potential Hypersurfaces: the function Potential hypersurfaces  are multidimensional functions that express potential energy of the whole system from the knowledge, at least in principle, from all particle coordinate and the kind of nuclei involved:

Potential Hypersurfaces: dimensions   The number of dimensions depends on the system and what is wanted or desired to model. This is a very key point and requires experience and professional abilities.

Potential Hypersurfaces: the point   The main problem of all modeling methods is to find the most appropriate function for the hypersurface of a given molecular system.

Potential Hypersurfaces: quantum nechanics   The only function that is consistent, from the theoretical point of view, to predict phenomena at the molecular level is that provided by quantum mechanics and related procedures

Potential Hypersurfaces: the quantum function   Therefore, a functional form that must be accepted as exact is:

Potential Hypersurfaces: quantum and classical modeling    The basic theory can help to solve the problem of the quantum hypersurface in two ways: Considering explicitly the electron density of molecules or quantum modeling. Treating the system as a set of classical bodies associated by springs and, therefore, governed by empirical force fields that simulate the quantum properties or classical modeling.

A quantum model of an inclusion compound

Calculated effects of inclusion

Classical modeling: an armonic oscillator   Almost all formulas are based on the harmonic oscillator potential, or Hook’s law:   Where r is the distance between two nuclei and k depends on the masses and the electronic surroundings of them.

Classical modeling: a posteriori methods   Methods are different because formulae and the way for obtaining parameters. All of them are typical a posteriori methods (MM4, AMBER, etc.).

Classical modeling: limitations for huge systems    In the case of very complex molecules, as proteins and all common biological hosts, even a comprehensive classical modeling results unaffordable.

Classical modeling: selecting coordinates   It means that the independent variables in the hypersurface   must be carefully selected, reducing substantially the dimensions of the system, and also the accuracy.

A model of a polypeptide

The computational paradox   The simplest methods are using less CPU time and, therefore, they can model larger polyatomic systems, depending on the available computational power. However, the predictive capacity is usually decreasing when methods are becoming simpler.

What is computational power?    If we consider computational power as the speed and efficiency for “crunching” numbers (even molecular graphics are intensive), the requirements for molecular modeling are unlimited.

What computational power depends on?    Computational power depends on : Processor speed able to perform as more as possible millions of operations per second. Storage capacity and speed of access in temporary memory. Storage capacity and speed of access in permanent memory.

Computational power for all    The advances in computer technology, that is nowadays very widely used by consumers, is giving us the possibility of performing very high level and state of the art scientific research with devices that could be bought in a supermarket, near to home.

Modeling hydrophobic molecular interactions 1 Amylose is a common adsorbent in chromatography and we could simulate it with a chain of three glucoses.

Modeling hydrophobic molecular interactions 2   If we explore the whole space of all possible associations between this trimer of amylose with benzyl benzoate, it was found by the  semiempirical method PM3 that one of the more favored associations is:

Modeling hydrophobic molecular interactions: View A

Modeling hydrophobic molecular interactions: View B

Modeling a drug in solution: 1    Amiphenazole is an activator of cell breathing that is banned by the  International Olympic Committee for athletes in competition.

Modeling a drug in solution: 2    Neutral amiphenazole: internal rotation four proton attracting sites tautomerism

Modeling a drug in solution: 3

Modeling a drug in solution: 4

Modeling a drug in solution: 5    It becomes clear that polar solvents influence equilibria in different ways depending on their protic character, and that environmental effects only taking into account the electrostatic field of the solvent are non comprehensive.

Modeling a drug in solution: 6

Modeling a drug in solution: 7

Docking approach: advantages and disadvantages The big advantage is that provides full understanding of the drug action mechanism at the very molecular level. The big disadvantage is that ignores all processes that must be followed by a molecule to arrive in a docking site if it is a drug.

Ligand approach 1 The ligand approach is applied if knowledge exists about the behavior as a drug of a certain pattern of molecular structure. This approach usually ignores the host structure, although research in this field gives always light inside drug action mechanisms.

Ligand approach 2

Ligand approach 3    The first step is always the selection of the kind of parameters to characterize the pattern molecular structure, as: Formula graphs Structural data Electron density related properties Shape related properties (dipole moments, integral surface, molecular volume, etc.)

Ligand approach 4   Those  parameters that serve for relating molecular properties are descriptors that could be correlated in many ways with pursued experimental facts. Then, mathematical tools allow predictions of biological activity and many other interesting applications.

QSAR 1 Quantitative structure-activity relationship (QSAR) methods became the more popular procedures to obtain useful results. The idea is to use a training set of molecules for obtaining confident mathematical relationships between the desired activity property and descriptors.

QSAR 2 It allows the further “synthesis in silico” of candidates, as more as possible, to calculate their expected behavior with the previously obtained relationships. Experimental work begins again with a reduced set of the more promising compounds, that must be tested in a traditional screening, saving much effort.

QSAR 3

A QSAR case 1 Penetration of antibiotics in cephalous rachid liquid (CRL) is expressed in terms of the ratio between the concentration in CRL with respect to blood in the presence of a certain infection:

A QSAR case 2 Confident experimental data of penetration pP of 18 antibiotics have been tested versus a large series of descriptors. After a careful search, a purely theoretical topologic descriptor (that depends on quantum chemical partial charges on atoms) W(q), the degrees of association with plasma proteins (DAPP), and an indicator variable (I) depending on the behavior of the antibiotic in the presence of H. influenzae or S. pneumoniae resulted significant.

A QSAR case 3 Correlation brings: with r=0.947 or with r=0.9748.

A QSAR case 4

3D-QSAR     A binding site at a receptor which ‘looks’ at a ligand would not see atoms and bonds. This is only a conventional language of chemists and biochemists. From a far distance, it would ‘feel’ the electrostatic potential of the molecule and, at a closer distance, the relatively hard body of the molecule with its charge distribution pattern at the solvent-accessible surface and the distribution of the most easily available electrons.

3D-QSAR: CoMFA    CoMFA is a comparative molecular field analysis that approaches active molecule comparisons by aligning them in space and by mapping their molecular fields to a 3D grid.    Partial least squares (PLS) and other data sampling and processing techniques are then used to generate QSAR equations.

A 3D-QSAR case: comparison between serotonin and LSD Serotonin is a neurotransmiter that docks in the so-called 5-HT2 receptor of presynaptic axon terminals. Lysergic Acid Diethylamide (LSD) is known as an ilegal drug that distort perceptions, and also docks in 5-HT2. Both of them are derivatives of triptophan

A 3D-QSAR case: formulas of serotonin and LSD

A 3D-QSAR case: molecular volumes of serotonin and LSD

A 3D-QSAR case: electrostatic potentials of serotonin and LSD

A 3D-QSAR case: HOMO's of serotonin and LSD

Ligand approach: advantages and disadvantages The big disadvantage is that considers all processes related to the site of docking for the drug molecule as a “black box”. The big advantage is that all processes followed by a drug molecule to arrive in a docking site are implicitly accounted. Therefore, the prediction of every case is reliable for practical success.

Software tools Scanners (data base search programs) Builders (denovo ligand design, optimizations of lead compounds, specific complementarity searching, visualizing) Specific task programs (mostly for numerically intensive calculations)

Conclusions?

Tools of computational chemistry are useful to model effects related to systems biology It is necessary to be familiar with methods, advantages and limitations It is necessary to be familiar with massive numeric treatment in computers The “instruments” could be those of any office

Revenues: It saves costly experimental work for: Many trial and error actions Understanding what is happening

This is a science where value is created by knowledge and procedures for handling information It is a good science at low cost…