Perspectives of Computational Chemistry

Fernando M.S. Silva Fernandes

Centre for Molecular Sciences and Materials (CCMM)

Department of Chemistry and Biochemistry (DQB), Faculty of Sciences (FC), University of Lisboa (UL)

Published in "Química, Boletim da Sociedade Portuguesa de Química, 123, 47-53, 2011"

 

 

Introduction

 

Laboratories and white gowns. Distillers, test tubes, burettes and pipettes. Acids, bases and salts. Fumaroles, colours and odours. Synthesis and analysis. Dyes, varnishes, detergents and plastics. Food and medicines. Biological cells, virus and bacteria. Electrodes, batteries and sensors. Formulae and molecular models. Art and geometry. Atoms and molecules in motion. Chemical, agricultural and pharmaceutical industries. Nuclear fissions and fusions. Stars, comets and planets. Internet, semi-conductors and computers. Agatha Christie and Conan Doyle.

 

These are realities where Chemistry is always present. Its vast domain is of crucial importance from atomic to galactic scales, contributing to unravel the Nature mechanisms, and to the health, welfare and progress of the ecosystem.

 

However, in some social circles still turns out, every now and then, the idea that chemists walk at the labs mixing reagents and using computers, in theoretical lucubrations and attempts to discover the "long-life elixir", or inventing bombs and other lethal products which will lead to the Humanity destruction. This image, drawn from alchemy, war and pollution, is unreal. Its rebuttal is urgent by fighting the subjacent scientific illiteracy.

 

Scientists are, in general, ethically and professionally responsible, though often over passed by thoughtless or ambitious political decisions.  Any scientific-technological activity that seeks innovation may have the "other side of the coin". Yet, only scientists and technicians, with their knowledge and methods, can correct less positive consequences that eventually turn out. Investing in science and technology (and humanities, of course), with justified criteria and giving effective opportunities to the newer generations, is the way to a healthy progress of Society.

 

In this article (a contribution to the celebrations of the International Year of Chemistry, IYC2011) we present essential aspects of Computational Chemistry, an interdisciplinary branch of Chemistry, serving all its specialities, and associated technologies and industries. The foundations were launched in the 1950's simultaneously with the implementation of the first computers. Nowadays, any curious and persistent person can learn it using a common personal computer connected to the Internet. It is, really, a challenge. There is who even claims that Computational Chemistry is an immensurable trunk belonging to the XXI and future centuries, from which some tiny pieces start falling out in the XX century.

 

Objectives, Methods and Programs

 

In what follows, some readers may find unfamiliar terms and techniques. To them we suggest the references [1-6], as sources of didactic material, software and programming techniques, covering most of the topics here addressed. Additionally, in the links to the research groups mentioned along the text there are other references related to the respective expertise.

Ultimately, Computational Chemistry consists of the development and usage of software dedicated to the resolution of chemical, biochemical, technological and industrial problems. The vast domain leads to the identification of some sub-branches dubbed Molecular Modelling (or Molecular Simulation), Chemometrics, Cheminformatics and Bioinformatics which, apart from considerable intersections, systematize objectives and specific methods:

  • Calculation of the properties of real or not yet synthesized molecules, and of molecular systems (solids, liquids, gases, plasmas, interfaces and biological organisms). The properties range from the electronic structures and conformations of isolated molecules to the different types of energy, dynamics and reactivity of molecular systems. The foundations are theories and models of classic and quantum mechanics, electromagnetism and statistical thermodynamics, whose typical computational methods are: ab initio and DFT, molecular mechanics and dynamics, Monte Carlo, including the ones of chemical kinetics, energy minimization, conformational and spectroscopic analysis, thermodynamic integration and perturbation, errors treatment and docking. These aspects are usually associated to the designation Molecular Modelling (or Molecular Simulation).
     

  • Analysis, treatment and retrieval of chemical information provided by laboratory experiments, instrumental/industrial monitoring and simulations, either in real-time or stored in data bases (typically with entry numbers of the order of several millions) to, for example: (a) prediction of spectra (NMR, Infrared, Mass, etc.) complementing the confirmation of chemical synthesis; (b) determination of quantitative structure-activity relationships (QSAR); (c) planning and assistance of automatic experimental techniques (such as combinatorial chemistry, high-throughput screening and flowshops) regarding the simultaneous synthesis of various products of chemical, agricultural and pharmaceutical interest; (d) determination of genetic sequences; (e) automatic classification of chemical and biochemical reactions; (f) pattern recognition; (h) quality control and instrumental calibration. Beyond molecular modelling techniques, more specific methods are used such as the analysis of principal components, multivariate optimization, artificial neural networks, genetic algorithms, cellular automata and expert systems which, in turn, are also used in many molecular modelling problems (energy minimization, conformational analysis, error estimates, regression, non-parametric fittings, design of new materials and drugs, etc.). These problems, at the boundaries of Chemistry, Statistics, Informatics and Artificial Intelligence, are within Chemometrics or Cheminformatics and Bioinformatics.

The requirements of numerical and symbolic calculations, and the storage, edition and visualization of huge amounts of chemical information, make the computer a key-instrument in Computational Chemistry. Like any machine it has to be instructed. Thus, programming is an indispensable resource, in languages as Fortran, C++, Java and Prolog. Actually, there exist several software packages, academic or commercial, with interactive molecular visualization (Gaussian, Gromacs, Towhee, Dock, Maddam, Spinus, ChemOffice, etc.) that can be used almost as "black-boxes", serving proficiently the diverse branches of Computational Chemistry. Yet, the research groups have frequently to develop their own programs, whenever the type of problems is not contemplated by the packages or these do not show the required reliability. Programming is a challenging intellectual pleasure (try it!) and, it is never too much stressing that chemists, biochemists, physicists and biologists are the main contributors to those software's, often available free of charge.

 

Depending on the kind of problems there are two essential types of computational programs:

 

For instance, the computation of electronic molecular structures or the molecular dynamics of a liquid are, generally, well defined problems in what concerns the input data and the expected results. The respective programs are of conventional type, that is, a set of elementary and sequential operations which process complete and precise data leading to unique results. They are strongly based in mathematic-numerical methods, more or less sophisticated, requiring always the ever increase of computer power.

 

Yet, there exist other chemical problems that can not be efficiently solved by such programs. The essential point is not more computational power and mathematical sophistication, but a different approach. Consider, for example, the determination of the number of isomers of C26H54O that are alcohols, the design of synthetic routes for new drugs and materials, the stereochemistry of proteins and the minimum energy conformations of other complex molecules, the genetic sequences, the selection of optimum parameters for instrumental analysis, the automatic control of large-scale industrial units, the dispersion of pollutants, and the query of chemical data bases. These cases identify the so-called combinatorial explosion, meaning that the number of hypothesis and details is so big that the eventual usage of conventional programs would not give responses within useful time, even having recourse to a supercomputer or sophisticated numerical techniques. Furthermore, either the inputs or the outputs of such cases possess, in general, intrinsic uncertainties implying that the adequate operations for their resolution do not have well defined sequences (as in the conventional algorithms), and the possibility of various answers with different confidence levels.

The approach to these problems has recourse to Artificial Intelligence methods, which attempt to simulate on a computer the intelligent reasoning, the genetic mechanisms, and the acquisition and manipulation of knowledge. One of the key-approximations is based on the fact that the brain, constituted by a network of neurons massively parallel, does not work sequentially. The neurons have, individually, a similar and simple function, but organised in parallel architectures possess the capacity of simultaneous and integrated processing, exchanging between them the information received from outside or generated internally, and manifesting, among others, two essential aspects of reasoning: "the intuitive jumps" and the ability to learn from training and experience.

 

Crossword puzzles illustrate the meaning of intuitive jumps. Indeed, humans do not execute an exhaustive search of all known words that can fit the indicated spaces. It often suffices a single letter to provoke an intuitive jump. A conventional algorithm would not have other alternative but an exhaustive search. On the other hand, the ability to drive a car can only be increased (that is, learned) by repeated training and memorizing. A conventional algorithm is unable to learn and memorize. Once written remains as such, unless the programmer alters it.

 

The programs of Artificial Intelligence (neural networks, genetic algorithms, expert systems, etc.) reproduce on a computer those capacities with considerable success, supporting the resolution of large-scale and complex problems.

For instance, the estimated number of molecules that may potentially act as drugs is of the order of 1040. From this astronomic number of possibilities only a tiny fraction will eventually be synthesized and, from this, just a smaller number will be submitted to high throughput screening for human tests. Any computational procedure designed to this end should be able to consider the huge number of potential drugs without inspecting effectively every molecular structure.

 

Furthermore, those programs allow the incorporation of an area of increasing chemical and technological importance: fuzzy logic. This logic shows that, after all, the formal rules of reasoning established by Aristotle are too rigid, not encompassing the majority of complex problems. Indeed, the answer to a question may not exactly be "yes" or "no" but "perhaps"; the colour of an object may not exactly be "white" or "black" but "grey", and within this there is an infinity of grades. Nowadays, it is common to find fuzzy processors in washing machines, cars and air conditioning. Such processors, contrary to the digital ones (of 0's or 1's), allow a continuous variation between 0 and 1, yielding substantial savings of energy and water.

 

Representative Topics

 

In Portuguese research groups at universities, institutes and industrial companies there is an intense and fruitful activity in the different sub-branches of Computational Chemistry. In what follows, we have selected applications, representative of the importance of this branch of Chemistry, developed in some of those groups. Yet, it should be stressed that, in the present article, we do not intend to review the national activity in this area, but just to exemplify a few computational aspects of pure and applied chemistry, biochemistry, agricultural chemistry and medicine.

In a future publication we shall present an exhaustive review on the Computational Chemistry in Portugal, resulting from an inquiry realized by the Portuguese Chemical Society. That would be unsuitable here considering the number of groups whose activity spreads from Minho to Algarve, Madeira and Azores. Thus, the groups not mentioned in the following selection will be referred to in the review, which we hope to be also presented in a web page in order to a direct linkage to the different group sites and permanent up-dating.

 

Nanoclusters and ionic liquids

 

Potassium chloride (KCl) is a salt (alkali halide) used, for example, in liquid fertilizers, food processing and medicine. It crystallises in a cubic geometry and melts at 1043 K (770 ºC). Its properties are generally determined (like the ones of any other substance) with samples that, though very small from a macroscopic standpoint, contain always a number of ions (K+ e Cl-) of the order of Avogadro's number(6.023x1023). Lord Kelvin conveyed the magnitude of that number in an appealing way: "Fill a glass of water whose molecules were somehow labelled. Pour the water in any ocean and let the molecules spread through all the world seas. Fill the glass again at any of the seas. About 100 molecules, initially labelled, shall be found".

 

How do behave microaggregates of KCl with a number of ions of the order of tens, hundreds or thousands? Will also present phase transitions and coexistence? At what temperatures? Potassium chloride can form glasses when the liquid is rapidly cooled. What about microaggregates? Will the laws of big numbers be applicable to systems with a very small number of particles? Note that the origin of any material resides in microaggregates with linear dimensions of nanometre order (10-9 m) dubbed nanoclusters. These questions are of fundamental importance.

 

Incidentally, Erwin Schrödinger in the famous classic "What is life?”, discusses the regularities observed in biological cells with fine reproducibility, resulting from the DNA which is constituted by a number of molecules very much smaller than Avogadro's number. Indeed, under a standpoint strictly macrophysical, only from macroscopic systems would emerge properties and laws with negligible fluctuations.

 

These aspects, are particularly suitable for computer simulation by the molecular dynamics method. Succinctly, a molecular dynamics program calculates the positions and velocities of the particles in every time instant (that is, their trajectories) using Newton's law (force = mass x acceleration). From the trajectories the computer can produce films, and determine multifarious properties such as temperature, pressure, kinetic, potential and free energies, diffusion coefficients, viscosities, thermal and electrical conductivities, spatial and temporal correlations and spectroscopic intensities.

The results obtained in a series of simulations on phase transitions and coexistence, illustrated by films produced by Pedro Rodrigues, are accessible in [7]. It is shown that nanoclusters present solid-liquid phase transitions whose melting points increase with the clusters size, approaching the melting temperatures of the bulk (clusters with only 512 ions have already melting points ~1000 K) and form glasses by fast cooling of the liquid. From ~1000 ions up, they also sustain phases coexistence (fig. 1). Will not nanoclusters be the "genes" of macro materials?

 

 

Figure 1.  Solid-liquid coexistence in a KCl nanocluster with ~5000 ions

 

By the way of materials, Rui Fartaria [8], for example, has carried out simulations on nano-composites of clay polymers, colloidal suspensions and liquid crystals. Moreover, at CICECO [9], simulations of different materials, concerned with investigations technology-oriented, have been realized.

 

In the context of ionic systems, the ionic liquids in particular, with important applications in cellulose processing and nuclear fuels, batteries, extraction of vegetable compounds regarding nutrition, cosmetics and pharmaceutical products, as well as in green chemistry, have had computational contributions of various groups, as the one of Canongia Lopes [10], in collaboration with experimental groups.

 

Adsorption and quantum simulation

 

Alcanethiols are easily adsorbed on gold electrodes trough the thiol group (-SH), due to the great affinity between sulphur and gold. After adsorption, from an ethanol solution, the molecules self-assemble on the gold surface forming dense and stable monolayers (fig. 2a). The electrochemical control of this process allows the chemical modification of the alkylic tails by introducing other molecules, as fullerene C60, giving rise, for example, to the development of different types of sensors.

On the other hand, the electro-oxidation of phenolic compounds on electrodes of noble metals is one possible technique to fight pollution.

 

(a)

(b)

 

Figure 2. (a) a step in the adsorption and self-assembling of decanethiol molecules (green) on gold electrodes from an ethanol

                solution; (b) potential energy surface of ethanol-gold.

 

The group of Fernando Fernandes [11] has carried out computer simulations of the mechanisms involved in those processes, in collaboration with experimental groups of interfacial electrochemistry, using quantum and Monte Carlo methods. The last is, in many aspects, equivalent to molecular dynamics but instead of generating deterministic trajectories, accordingly Newton's law, does a stochastic sampling (a stochastic process characterizes an evolution subjected to the laws of probability) of the positions and/or velocities of the particles. These problems, and others in the domains of molecular interactions, reactivity and materials science, have also had contributions, for instance, from Ferreira Gomes's group [12].

 

It should be emphasized that Quantum Mechanics is always subjacent to Computational Chemistry methods, though often in a non explicit form. Atomic, molecular and nuclear structures, spectroscopy, approximate force fields for molecular dynamics and Monte Carlo, and potential energy surfaces (fig. 2b) for the study of interfaces and chemical reactions, have invariably a quantum support. In this context, the group of António Varandas [13] has made seminal developments to quantum chemistry, and respective computational methods, with diverse applications in particular to atmospheric chemistry. Among others, the groups of Benedito Cabral [14] and of Prates Ramalho [15] have also carried out research on quantum simulation (solids, fluids and clusters).

 

Neural networks and prediction of NMR spectra

 

Nuclear Magnetic Resonance (NMR) is a spectrometric technique applied in all branches of chemistry, in industry and medicine. It is based on the fact that the protons and neutrons of the atomic nuclei have spins (rotational motions, roughly) which originate intrinsic magnetic fields that can interact with applied external fields. The electrons surrounding the atoms in a molecule shield the nuclei to the external fields. This shielding effect turns NMR useful for the identification of molecular structures, since the shielding depends on the electronic distributions which determine the different chemical bonds. Thus, different nuclear resonance frequencies are expected corresponding to the various neighbourhoods of the nuclei in a given molecule. Such differences, called chemical shifts, are defined relatively to a standard, for example, tetramethylsilane (TMS). The technique takes into account several nuclei such as 1H (proton), 13C, 15N, 19F and 31P.

 

When a new molecule is synthesized its confirmation is necessary. So, it is of the utmost importance to predict the spectrum, of NMR for instance, and afterwards to compare it with the experimental one. Such is possible with neural networks programs that learn from different types of spectra for known molecules. Once trained, they can predict, with remarkable reliability, the spectra of new molecules. An example is the program SPINUS developed in the group of Aires de Sousa [16] for proton-NMR spectra, accessible in [17]. All the user has to do is to draw, with the graphical interface included, the required molecular structure and, then, a single click gives the spectrum in terms of chemical shifts (fig. 3). Each peak corresponds to a hydrogen atom in the compound having different neighbourhoods. The spin-spin couplings, responsible for the spectra fine structure, are also determined.

 

 

 

Figure 3. Interface of program SPINUS sowing the structures (2D and 3D) of a molecule and the respective proton-NMR spectrum (with details of fine structure)

 

Another application concerns structural elucidation, that is, deducing the structure of a new compound from experimental spectroscopic data. The programs to this end incorporate predictions of spectra, like the ones of SPINUS, to filter among the possible candidate-structures.

 

The referred to group has also developed other applications of artificial intelligence methods, such as the geographic distribution of petroleum samples, classification of enzymatic reactions at genomic scale and, particularly Diogo Latino [18], non-parametric fittings of intermolecular potential energy surfaces. 

 

Dynamics of proteins, dendrimers and lipids

 

Molecular visualization is an indispensable resource, in special for crucial molecules in the cellular structures and biologic processes. It complements the calculations, detects details that otherwise may remain hidden and suggests new experiments. For example:

 

Lysozyme is an enzymatic protein (exists in the hen egg white, human milk, saliva, etc.) of the natural immune system. Being a natural form of protection to bacteria as Salmonella, E.coli and Pseudomonas, its deficiency in the nourishment, specially of newborns, has been associated to bronchitis, pneumonia and diarroea.

 

The interactions between the functional groups of the phospholipid cardiolipin  (CL) result in molecular conformations suggested as determinant for the good performance of the internal mitochondrial membrane, being the deficiency of CL associated to Barth's syndrome.

 

Dendrimers are polymeric molecules whose architecture is characterized by successive branches, often approximately spherical, conferring to them the designations of 1st, 2nd, 3rd... generation dendrimers. There exist diverse dendrimers families which, though with the same architecture, are distinguished by the type of monomers (the basic units of a polymer). Being polymeric molecules they have similarities with proteins, and in the case of peptidic dendrimers (a family with an increasing development since 2005) the respective monomers are amino acids as in proteins. Relatively to these they show a considerable molecular flexibility.

The applications of dendrimers are multifarious and promising in biochemistry, nanomedicine and industry. The architecture and chemical properties of the molecules are particularly suitable to, for example, encapsulate (fig. 4) drugs and genes carrying them to cancerous tissues and to DNA of biological cells. Besides the delicate strategy of the technique, it facilitates the attenuation of the toxicity of such treatments.

 

Films from molecular dynamics simulations of lysozyme, cardiolipine (in different situations) and a peptidic dendrimer (showing its molecular flexibility relatively to proteins), produced by the group of Miguel Machuqueiro, are accessible in [19].

 

 

Figure 4. Sketch of a dentritic "box" of 3rd generation encapsulating two molecules

 

Modelling and dynamics of proteins, dendrimers and akin aspects have had other contributions, for example, from the groups of Cláudio Soares [20] and António Baptista [21]

 

Enzymatic activity and drug design

 

Enzymes are proteins that catalyse chemical reactions. Most of biological processes need enzymes whose activity is affected by other molecules, which can act as inhibitors or activators. Many drugs and poisons are enzymatic inhibitors.

 

The reactional mechanisms, that is, the way how the molecular chemical bonds break or form, depend directly of the electronic structure of the molecules. A determinant step in any reaction is the formation of the so-called activated complex, i.e., an aggregate between the reagent molecules with adequate conformations and relative orientations for obtaining the products. Quantum and molecular docking computational methods are the first steps to clarify the enzymatic mechanisms involved in drug design.

 

For instance, the virus HIV-1 of AIDS has an enzymatic activity of extreme complexity. It produces three enzymes: the integrase that commands the integration of its genetic material into the infected cells, the reverse transcriptase that converts the single viral chain of RNA in the double DNA chain, ready to be inserted in the human DNA, and the protease that activates the destruction of proteins synthesized in the cells. The design of inhibitors (fig. 5) to fight this disease has had contributions of the group of Maria João Ramos. Some aspects, accessible in [22], are illustrated by films.

 

 

Figure 5. A step in the association of a molecule with therateupical interest to HIV-1 protease

 

Further computational developments within Biochemistry, and Medicinal and Pharmaceutical Chemistry, have also been achieved, for example, at the groups of Nuno Micaelo [23], Vitor Félix [24], Daniel dos Santos [25] and Nuno Palma [26, 27]

 

Metrology and chemometrics

 

Using computers and other instruments, or even a single sheet of paper and pencil, the researchers and technicians determine values for various properties of systems and processes. That is, they always perform measurements, whose reliability, correct presentation and comparisons only can be guaranteed through the definition of international system of units, evaluation of uncertainties and errors, and the calibration of the instruments. On the other hand, health and public and environmental safety, demand strict legal norms that regulate the instrumental calibrations, the formulation of products and the industrial processes. The non- fulfilment of such norms either, e.g. in the calibration of aircraft altimeters, or in the formulation of medicines or pesticides is disastrous as, unfortunately and often, has been observed.

 

These problems, of every scientific and technological field, are the domain of Metrology (not meteorology, note) that uses, in particular, computational methods of Chemometrics.

 

The report and quantification of the chemical species involved in multiple chemical equilibria, and the modelling of the metrological performance of chemical measurements able to identify the origin of interlaboratory discrepancies, in the conformity assessment of food in complex matrices (fig. 6) with legislation, are examples of situations with strong social and economical impact which benefit from computational tools developed or applied to such purpose. The group of Filomena Camões [28], for instance, has contributed to this area.

 

 

Figure 6. Model of the compatibility of pesticide residues measurements in food, considering

                                different conventions for reporting the results and estimations of measurement uncertainties.

 

In the context of metrology for instrumental calibration, in particular, the group of Nieto de Castro [29] is another example. 

 

Final remarks

 

A popular text should avoid technicalities as far as possible, keeping, however, the scientific rigour. With this in mind, we have outlined the objectives, methods and types of programs of Computational Chemistry, and selected some examples representative of the activity going on in Portugal.

 

In the present context, we should underline the prizes Alberto Romão Dias and Vicente Seabra awarded, by the Portuguese Chemical Society in 2011 and 2010, respectively to Maria José Calhorda [30], University of Lisboa, and to José Richard Gomes [31], University of Aveiro, in whose research groups Computational Chemistry plays a fundamental role.

 

Communicating science to non-professionals is a civic duty, in particular of scientists. We believe that is one of the ways to stimulate the newer generations, fight the scientific illiteracy, and convey to the society the successes (and also the failures, of course) of the permanent endeavour of Science to contribute to the progress and welfare of societies. Even for investigations classified as "top-secret" there are always ways to divulge certain general aspects and techniques, the last often very well-known, maintaining the "suitable level of unawareness" of others.

 

We finish off recalling what was said in the Introduction. Any curious and persistent person is able to learn Computational Chemistry, using a common personal computer connected to the Internet. And, it is not exaggeration stating that not always is necessary be an expert to produce excellent scientific ideas. Reading a little of History of Science confirms it. Here is the challenge, dear readers.

 

Acknowledgements

 

To colleagues Pedro Fernandes (Fac. Sciences, Univ. Porto), João Aires de Sousa and Diogo Latino (Fac. Sciences and Tech., Univ. Nova de Lisboa), Filomena Camões, Ricardo Bettencourt, Miguel Machuqueiro and Pedro Rodrigues (Fac. Sciences, Univ. Lisboa) for providing some figures and films as well as comments and suggestions to the text. To the referee of this article for the critics and opportune suggestions.

 

References

 

[1]  Open Source Physics (http://www.opensourcephysics.org/)  
[2]  
Wiki Ciências (http://wikiciencias.casadasciencias.org/index.php/)

[3]  "Using Artificial Intelligence in Chemistry and Biology, A Practical Guide", Hugh Cartwright, CRC Press, 2008.

[4]  "Cinquentenário da Simulação Computacional em Mecânica Estatística I. Os primeiros passos",
       F.Fernandes, Química, Bol. SPQ, 90 (2003) 39-43.

[5]  "Cinquentenário da Simulação Computacional em Mecânica Estatística II. Desenvolvimento e aplicações

       fundamentais", F.Fernandes, Química, Bol. SPQ, 93 (2004) 49-60.

[6]  "Monte Carlo em Mecânica Estatística. Uma explicação simples com argumentos cinéticos",
       F.Fernandes, Revista Ciência, AEFCUL, Série VII, 2 (2001) 31-39.

[7]   Pedro Rodrigues (http://elixir.dqb.fc.ul.pt/~fsilva/clusters_IYC2011)

[8]   Rui Fartaria (http://elixir.dqb.fc.ul.pt/~fartaria/home/)

[9]   CICECO (http://www.ciceco.ua.pt/)

[10] Canongia Lopes (http://cqe.ist.utl.pt/personal_pages/pages/jose_nuno_lopes.php/)

[11] Fernando Fernandes (http://elixir.dqb.fc.ul.pt/~fsilva/home/)

[12] Ferreira Gomes (http://www.fc.up.pt/pessoas/jfgomes)

[13] António Varandas (http://www.uc.pt/uid/tcc)

[14] Benedito Cabral (http://gfm.cii.fc.ul.pt/people/bjcabral/)

[15] Prates Ramalho (http://home.uevora.pt/~jpcar/)

[16] Aires de Sousa (http://joao.airesdesousa.com)

[17] SPINUS (http://www.dq.fct.unl.pt/spinus)

[18] Diogo Latino (http://elixir.dqb.fc.ul.pt/~latino/)

[19] Miguel Machuqueiro (http://intheochem.fc.ul.pt/research/artwork.html)

[20] Cláudio Soares (http://www.itqb.unl.pt/~claudio/)

[21] António Baptista (http://www.itqb.unl.pt/~baptista/)

[22] Maria João Ramos (http://www2.fc.up.pt/A-quimica-do-dia-a-dia/)

[23] Nuno Micaelo (http://www.simulation.quimica.uminho.pt/)

[24] Vitor Félix (http://molecular-modeling.dq.ua.pt/)

[25] Daniel dos Santos (http://www.imed.ul.pt/portal/index.php?option=com_content&view=article&id=215&Itemid=80)

[26] Nuno Palma (http://www.linkedin.com/in/nunopalma64)

[27] Bial (http://www.bial.com/pt/)

[28] Filomena Camões (http://www.dqb.fc.ul.pt/pessoal/mfcamoes.php)

[29] Nieto de Castro (http://www.dqb.fc.ul.pt/pessoal/cacastro.php)

[30] Maria José Calhorda (http://intheochem.fc.ul.pt/)

[31] José Richard Gomes (http://sweet.ua.pt/~f3963/)