Applications of Artificial Intelligence in Chemistry details the various applications of artificial intelligence in the chemical science fields. Artificial intelligence is not just about making machines think; it is also a powerful problem-solving tool. Many scientific problems can be solved only with difficulty using conventional methods, yet these same problems may be ideally suited to attack using artificial intelligence. The chapters cover artificial intelligence, artificial neural networks, expert systems, and genetic algorithms.
Gillian Pocock and Hugh M
This chapter explores the extensive potential of artificial intelligence (AI), which has been recognized and harnessed by computer scientists, including physics and life sciences. It analyses how AI is applied to those areas of chemistry that offer great scope for investigation by intelligent methods. It also defines AI as an attempt to replicate intelligent reasoning through ‘machines’, which the chapter uses in a sense that is much broader than its everyday meaning. It highlights the differences between the way scientific problems are tackled using conventional methods and the way they are solved using the alternative methods of AI. It discusses how computers can be persuaded to act intelligently and solve problems that may resist solution using ‘dumb’ methods.
Artificial neural networks
This chapter covers artificial neural networks, which are an example of a connectionist model with many logical units that are connected in a network. It reviews the way in which the brain is believed to function, as the design of the artificial neural networks is inspired by the structure of that organ. It also explains that the neuron is the fundamental processing unit in a typical human brain. The chapter talks about the behaviour of the neuron as a threshold device that has a step-like shape known as a step function or Heaviside function. The chapter points out the crucial feature of the behaviour of the brain, which is that it does not need to be taught how to learn. It observes that artificial neural network programs are multi-purpose as they can solve problems by using spectral interpretation, propositional logic, image analysis, or fingerprint interpretation.
Chemistry and the Internet
This chapter discusses how chemistry has benefited from the internet as there are countless sources of chemical information available online. It explains that the graphical aspects of the internet are of great advantage for chemistry, as molecules and other things can be displayed in three dimensions or animated in ways not possible with standard paper publications. However, the evolution of protocols and formats of the data needed to produce complex chemical graphics also pose challenges in that a particular computer that does not have the right software installed may fail to display the graphics. The chapter also analyses the evolving nature of the internet and argues that this creates a big problem as publications become outdated as soon as they have been written. It discusses the vast amounts of information available on the internet on all aspects of chemistry.
Computation and Computers in Chemistry
This chapter provides an overview of computation and computers in chemistry. Devising an experiment or analysing its results usually involves some calculation or other. A more fundamental role for computation is to work out the details of the prediction of a particular chemical theory. Very often, this requires so many calculations that they cannot be done by hand or with a calculator—a computer is needed instead. In modern chemical research, computation is used for such diverse aims as studying the diffusion of molecules through a membrane, for exploring the motions of one part of a protein with respect to the others, for predicting the features in an electronic absorption spectrum, or for trying to understand the mechanism of a chemical reaction. This book explores how the corresponding methods work and introduces the computational techniques used to carry out such calculations.
H. Grant Guy and Richards W. Graham
Computational Chemistry starts by arguing that the uses of computers in chemistry are many and varied. This ranges from the modelling of solid state systems to the design of complex molecules which can be used as drugs. This text introduces the many methods currently used by practising computational chemists and shows the value of computers in modern chemical research. The text describes the various computational techniques available and explains how they can be applied to single molecules, to assemblies of molecules, and to molecules undergoing reaction. An introductory chapter outlines the hardware and software available, and looks at some applications and developments. Subsequent chapters cover quantum mechanics, molecular mechanics, statistical mechanics, the modelling of biomolecules, and drug design. Whilst emphasizing the use of computers to model biological systems, the chapters explain how the methods can be applied to a whole range of chemical problems.
This chapter focuses on computational chemistry, which is explained as using computers to calculate chemistry, such as molecular properties, reaction rate, fundamental properties of a system, or theoretical models as basis for calculating other properties. It stresses how computational chemistry has become an important branch of chemistry as the power of computers increases with faster and more complex calculations. The chapter also discusses program packages which have been developed to allow non-programmers to perform complex calculations. The chapter describes two different forms of packages: ones that are designed for a particular application and others which are more general in nature. It discusses the Gaussian commercial package, which is used to calculate molecular parameters from a theoretical basis, as an example of the former.
Computational Chemistry begins with an introduction to computation and the use of computers in chemistry. Next it turns to quantum chemistry and quantum chemical methods. The chapter that follows examines molecular mechanical methods. There follows a chapter on geometry optimization. The next couple of chapters look at dynamics and rate constants and equilibria. The last main chapter is about hybrid and multi-scale methods.
Computers in Chemistry introduces the use of computer technology in the chemical sciences. Computers have become an integral part of chemistry. Virtually all modern scientific instrumentation contains some form of computer and, indeed, the operation of many instruments has become so complex that it is impossible without some degree of computer control. It is very important for the modern student of chemistry to have at least a basic knowledge of computers, and the deeper that knowledge is, the better use will be made of the techniques available.
The object of the preceding chapters was to provide an introduction to many of the methods currently used by practising computational chemists; in no way was it intended as a ‘how to do it’ manual. Although the emphasis has been on computer modelling of biological...
This concluding chapter describes how computational methods are used in a wide variety of ways to provide a better understanding of chemical phenomena. Computers are more and more powerful, new and more accurate approximations are being developed, and algorithms are becoming increasingly efficient, so that overall it is now easier than ever before to perform calculations that provide accurate models of experimental systems in all areas of chemistry. These developments are expected to continue in coming years, making this an anticipated golden age for computational chemistry. The chapter then considers some general aspects that should be borne in mind when designing a computational study. Careful design of a project requires that the aims be well defined, and the uncertainties well understood.
This chapter focuses on dynamics methods. Especially for complex systems, the ‘static’ inspection of features of the potential energy surface is not sufficient to predict the behaviour of the system described by it. Simulation techniques sample the range of structures that are visited under particular experimental conditions. Molecular dynamics simulations perform this sampling by applying Newton’s laws of motion to the atoms. On the other hand, Monte Carlo simulations use random changes of structure, with an algorithm for accepting or rejecting given changes that results in an overall Boltzmann distribution. Sampled structures can be analysed in terms of average properties, for example using radial distribution functions. The most common application of dynamical simulation methods is the study of biomolecules, their structural variability, and their interactions with other molecules.
This chapter describes the ‘expert’ as a valuable commodity in the field of science which requires a high level of technical understanding. It focuses on an expert system, which is a computerized clone of a human expert and contains a fund of knowledge relating to some specific, well-defined area. It also analyses how a computer can reproduce many of the logical deductions of a human expert and offer scientific advice of a quality rivalling that which a human might provide. The chapter outlines how expert systems operate and how they can be applied to science. It discusses how computers can reason and draw conclusions and mimic what is commonly known as ‘thinking’.
This chapter talks about genetic algorithms (GA), which are considered an optimization technique. This is, in other words, an intelligent way to search for the optimum solution to a problem hidden in a wealth of poorer ones. It explains that a GA works with a ‘population’ of individuals. It also clarifies how the individuals ‘mate’ with each other, ‘mutate’, and ‘reproduce’ in order to evolve through successive generations toward an optimum solution. The chapter discusses what characteristic of evolution acquires the ability to provide the inspiration for solving numerical problems and how its power can be harnessed in a most effective and intriguing way. It mentions the similarities of GA with other optimization and search methods in which any computational task involves mating, mutation, and reproduction.
This chapter details the features of potential energy surfaces, with a focus on energy minima and saddlepoints or transition states. It also describes methods to explore potential energy surfaces in a systematic way through geometry optimization. Efficient optimization methods are based on calculation of the potential energy, its first derivative with respect to the positions of the atoms (the gradient), and the second derivative (the Hessian matrix). Geometry optimization is frequently used in quantum chemical studies; predicted molecular structures usually agree very well with experiments. The curvature of the potential energy surface around the optimum structure can be used to predict vibrational frequencies. As well as minima, optimization can also be used to locate transition states and reaction paths.
This chapter describes what a computer is physically made up of. It reviews the basic elements of a computer, such as the central processor unit (CPU), the memory that stores results and the controlling program, the input and output (I/O) devices that communicate with the outside world, and the buses that provide communication among the various elements. It also refers to the specific memory location which is controlled by the address bus and data which are read or written on the data bus. The chapter describes the arithmetic-logical unit (ALU) along with its associated registers and control unit found in the CPU. It talks about the architecture of a CPU, which includes the layout of the registers, ALUs, and control circuitry within the chip.
Hybrid and Multi-Scale Methods
This chapter addresses hybrid and multi-scale methods. Chemistry of a ‘core’ system is frequently perturbed by the ‘environment’. If the perturbation is weak, as is often the case, it is acceptable to perform calculations on the core part only. The effects of the environment can in many cases be described using continuum models, which can conveniently be coupled to quantum chemical calculations. It is also possible to devise hybrid methods, in which the atoms making up the core and the environment in the model are treated at different levels of theory. One very popular family of hybrid methods treats the core quantum mechanically (QM) and the environment with molecular mechanics (MM), and these methods are referred to as QM/MM methods.
Interfacing computers to experiments
This chapter introduces techniques on getting experimental information into a computer and causing a computer to control the experiment. It explains that an analogue to digital converter (ADC) is a device that is used to measure voltage, while the digital to analogue converter (DAC) is a device used to generate a specific voltage. It also describes the output of an ADC in a computer system. This is designed to reflect one or more memory locations on the microprocessor bus. The chapter discusses the parallel input and output (PIO) device, which allows a single bit or group of bits to be output or input. It also senses whether a signal is on or off. It highlights the principle of DAC, wherein the device produces a voltage proportional to binary data.
This introductory chapter provides an overview of computational chemistry, which is one of the fastest growing areas of chemistry. Although there are specialists in the field, increasingly the techniques are applied by experimental chemists using the ever-growing power of ever-cheaper computers. Ultimately, computational chemistry involves taking known theory and developing the computer software to solve chemical problems. The chapter then looks at the computer hardware and software that many computational chemists use. The main classes of problems which can be resolved by computer include single molecule calculations, assemblies of molecules, and reactions of molecules. Every time a significant advance has been made in computing technology or in application, computational chemists have seized the opportunity and incorporated it into their own field. The chapter considers two examples to illustrate this: computer graphics and neural networks.