CHAPTER ONE
1.0 INTRODUCTION
1.1 Data Mining
Data mining involves the use of sophisticated data analysis tools to discover previously unknown, valid patterns and relationships in large datasets. (Adriaans and Zantinge,1999).
Data mining (sometimes called data or knowledge discovery) is the process of analyzing data from different perspectives and summarizing it into useful information which was proposed by (Zhu and Cao, 2003). The overall goal of the data mining process is to extract information from a data set and transform it into an understandable structure for further use (ACM SIKKDD, 2006). Data mining software is one of a number of analytical tools for analyzing data./ It allows users to analyze data from many different dimensions, categorize it and summarize the relationships indentified.
Technically, data mining is the process of finding correlation or pattern among dozens of fields in large relational databases. Data mining consists of more than collecting and managing data, it also includes analysis and prediction. (Wilson et al., 2003) refer to data mining as utilizations of statistical techniques within the knowledge discovery process. Data mining is used for a variety of purpose in both the private and public sectors. In public sector, data mining applications were initially used as a means to detect fraud and waste. It has been reported that data mining has helped the Federal Government recover millions of dollars in fraudulent Medicare payments (Cashlink, 2000). Another example is the Federal Aviation Administration, which uses data mining to review plane crash data to recognize common and recommend precautionary measures.
1.2 Types of Data Mining
Classification
Clustering
Classification
Classification is a learning function that maps (classify) a data item into one of several predefined classes (Weiss and Kulikowski, 1991). Examples of classification methods used as part of knowledge discovery applications include the classification of trends in financial markets and the automated identification of objects of interest in large image databases.
Clustering
Clustering is a common descriptive task where one seeks to identify a finite set of categories or clusters to describe the data. The categories can be mutually exclusive and exhaustive or consist of a richer representation, such as hierarchical or overlapping categories. Examples of clustering applications in a knowledge discovery context include discovering homogenous sub- populations for consumers in marketing databases and identifying subcategories of spectra from infrared sky measurements. (Wilson et al 2003).
1.3 Data Mining Tools
Data mining tools predict future trends and behaviour allowing business to make proactive analyses offered by data mining and move beyond the analyses of past events provided by retrospective tools typical of decision support system. Data mining tools can analyze massive database to deliver answers to questions
Most data mining tools can be classified into three categories:
· Traditional data mining tool
· Dashboard data mining tool
· Text- mining data tool
Traditional Data Mining Tools: It helps companies to establish data patterns and trends by using a number of complex algorithms and techniques. Some of these tools are installed on the desktop to monitor the data and highlight the trends and others capture information residing outside a data house. The majority are available in both windows and UNIX versions, although some specialize in one operating system only. In additional, while some may concentrate on one database type, most will be able to handle any data using online analytical processing or a similar technology. (Zhu & Cao, 2003).
Dashboard Mining Tools: It is installed in computers to monitor information in a database, dashboards reflect data charges and up dates on screen often in the form of a chart or table enabling the user to see how the business is performing. Historical data also can be referenced, enabling the user to see where things have changed. This functionality makes dashboards easy to use and particularly appealing to managers who wish to have an overview of the company’s performance. (Zhu & Cao, 2003).
Text-Mining Tools: It is the ability to mine data from different kinds of text from Microsoft Word and Acrobat PDF document to simple text files. These tools scan content and convert the selected data into a format that is compatible with the tools database, thus providing users with an easy and convenient way of accessing data without the need to open different applications scanned content can be in structured (i.e., information is scattered almost randomly across the document including e-mail, internet pages, audio and video data) or structured (i.e, the data’s form and purpose is known such as content found in database capturing these inputs can provide organizations with a wealth of information that can be mined to discover trends, concepts and attitudes). (Zhu & Cao, 2003).
1.4 Data Mining Techniques
Data mining techniques can yield the benefits of automation on existing software and hardware platforms, and can implement on new systems as existing platforms are upgraded and a new product developed.
The most commonly used techniques includes artificial neural networks, decision trees, nearest neighbor method, genetic algorithms, rule induction, data visualization, self-organizing map, and short time series expression miner.
Decision Trees: Are tree-shaped structure that represents decision sets. These decisions generate rules classification of a dataset. Specific decision tree methods include classification and regression trees (CART) and Chi-square Automatic Interaction Detection (CHAID).
In Zhu et al., (2003), CART and CHAID are decision tree techniques used for classification of a dataset. They provide a set of rules that you can apply to a new data set to predict which records will have a given outcome. CART segments a dataset, by creating 2-way splits while CHAID segments using chi-square tests create multi-way splits. CART typically requires less data preparation than CHAID.
Genetic Algorithms: According to Zhu et al., (2003), genetic algorithm is optimization techniques that use processes such as genetic combination, mutation and natural selection in a design based on the concept of natural evolution.
Nearest Neighbour Method: It classifies dataset records based on similar data in a historical dataset. A technique that classifies each record in a dataset based on a combination of the classes of the k-record(s) most similar to it in a historical dataset (Where K1) sometimes called the k-nearest neighbor technique.
Data Visualization: The visual interpretation of complex relationship in multidimensional data graphics tools are used to illustrate data relationships.
Rule Induction: Induction is the process in which a computer program generates rules from sample cases or other data. A rule induction system is a given example of a problem where the outcome is known. These examples are called the training set. At the basis of a rule induction system is an algorithm which is used to induce the rules from the examples. An example is the ID 3 algorithm (Quinlan, 1986). However, all induction algorithms walk in similar fashion and are all based in some way on statistical analysis of the training set. Advantages of machine induction include:
· The ability to deduce new knowledge. It may be possible to list all the factors influencing a decision without understanding their impacts.
· Once rules have been generated, they can be reviewed and modified by the domain expert.
1.5 Self-Organizing Map
A self-organizing map (SOM) or self organizing feature map (SOFM) is a type of artificial neural network that is trained using unsupervised learning to produce on low dimension discretized representation of the input space of the training samples, called a map. Self-organizing maps are different from other artificial neural network in the sense that they use a neighbourhood function to preserve the topological properties of the input space. This makes SOMS useful for visualizing low-dimensional views of high-dimensional data, taken to the multi-dimensional scaling; the model was first described as an artificial neural network by the professor Teuvo Kohonen, and is sometimes called a Kohonen map, (Kohonen & Honkela, 2007).
A self- organizing map consists of components called nodes or neurons. Associated with each node is a weight vector and a position in the map space. It describes a mapping from a higher dimensional input space to a lower dimensional map space, while it is typical to consider this type of network structure as related to feed forward networks where the nodes are visualized as being attached, this type of architecture is where the nodes are visualized as being attached, this type of architecture is fundamentally different in arrangement and motivation. The self-organizing map is especially suitable for data survey because it has prominent visualization properties (Kohomen, 1995).
1.6 Short Time Series Expression Miner
The advert of the microarray has been hailed as a revolution in molecular biology (Cheung et al., 1999). It is the era of high throughput gene expression analysis, allowing researchers to interrogate the expression levels of many genes at once.
Two of the most recent algorithms are short time series expression miner (STEM) (Ernst and Bar-Joseph, 2005) and selection of informative genes via Symbolic Hashing of Time Series (SLINGSHORS) (Yang et al., 2006). These algorithms both perform fine grained clustering through the use of a hashing function which assigns similarly shaped expression profiles to a specific motif, and work under the assumption that highly populated motifs are more relevant to an organisms underlying response to an external stimulus than motifs which were sparsely populated.
1.7 Clustering
A clustering mean partitioning a data set into a set of clusters, each data sample belongs to exactly one cluster. Clustering can also be based on mixture models (Mclahlan and Basford, 1987). A widely adopted definition of clustering is a partitioning that minimizes distance within and maximizes distances between clusters. Clustering is the far most used method in gene expression analysis. A hierarchical cluster is currently the most frequently applied method is gene expression analysis.
1.8 Neural Network
An artificial neural network (ANN), usually called neural network (NN) is a mathematical model or computational model that is inspired by the structure or functional aspects of biological neural networks. An artificial neural network is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information. A network consists of an interconnected group of artificial neurons, and it processes, information using connectionist approach to computation. In most cases, an ANN is an adaptive system that changes its structure based on external or internal information that flows through the networks are non-linear statistical data modeling tools. They are usually used to model complex relation between inputs and outputs or do find patterns in data. (Bertsekas & Tsitslklis, 1996).
Neutral network models in artificial are usually referred to as artificial neutral networks (ANNS). The word network in the term ‘artificial’ neural network’ refers to the inter-connections between the neurons in the different layers of each system. An example system has three layers. The first layer has input neurons, and then via more synapses to the third layer of output neurons.
An ANN is typically defined by three types of parameters
· The interconnection patterns between different layers of neurons.
· The learning process of updating the weights of the inter- connections.
· The activation function that converts a neutron’s weighted input to its output activation
Mathematically, a neuron’s network function f(x) is defined as a composition of other function gi(x), which can further be defined as a composition of other functions. A widely used type of composition is the non-linear weighted sum, where f(x)=k (Eiwigi(x), where k (commonly referred to as the activation function (the machine learning dictionary) is some predefined function, such as the hyperbolic tangent (Kohenen & Honkela,2007).
Neural networks, as used in artificial intelligence, have traditionally been viewed as simplified models of neural processing in the brain, even though the relation between this model and brain biological architecture is debated, as it is not clear to what degree artificial. Neural networks mirror brain function (Russe and Ingrid, 2012). Neural coding is concerned with how sensory and other information is represented in the brain by neurons. The main goal of studying neural coding is to characterize the relation between the stimulus and the individual or ensemble neuronal responses and the relationship among electrical activity of the neurons in the ensemble (Brown et al., 2004). It is thought that neurons can encode both digital and analog information (Thorpe, 1990).
1.9 Employing Artificial Neural Networks
The greatest advantage of ANN is their ability to be used as an arbitrary function approximation mechanism that learns from observed data. However, using them is not so straight forward and a relatively good understanding of the underlying theory is essential. This will depend on the data representation and the application. Overly complex models tend to lead to problems with learning (Kohenen & Honkela, 2007).
With the correct implementation, ANN can be used naturally in online learning and large data set application. Their simple implementation and the structure allows for fast, parallel implementations in hardware.
Applications
The utility of artificial neural network models lies in the fact that they can be used to infer a function from observation. This is particularly useful in application where the complexity of the data or task makes the design of such a function by hand impractical (Kohenen & Honkela, 2007).
Real-Life Applications
The tasks artificial neural networks are applied to tend to fall within the following broad categories:
· Function approximation, or regression analysis, including time series prediction, fitness, approximation and modeling.
· Classification including pattern sequence recognition, novely detection and sequential decision making.
· Data processing, including filtering, clustering, blind source separation and compression.
· Robotics, including directing manipulators, computer numerical control.
Application areas include system identification and control vehicle control, process control, natural resources management), quantum chemistry, (Balabin et al, 2009). Game-playing and decisions making (backgammon, chess, pokar), patterns recognition (radar systems, face identification, object recognition and more), sequence recognition (gesture, speech, handwritten text recognition), medical diagnosis, financial applications (automated trading systems), data mining (or knowledge discovery in database, “KDD”), visualization and e- mail spam filtering.
Artificial neural networks have also been used to diagnose several concerns. An ANN based hybrid lung cancer detection system named HLND improves the accuracy of diagnosis, N. These networks have also been used to take specific models taken from a large group of patients compared to information of one given patient. The models do not depend on assumption about correlations of different variables. Colorectal cancer has also been predict the outcome for a patient with colorectal cancer with a lot more accuracy than current clinical methods after training, the networks could predict multiple.
Types of Artificial Neural Network
Artificial neural network types vary from those with only one two layers of single direction logic, to complicated multi-input many direction feedbacks loop and layers. On the whole, these systems use algorithms in theory programming to determine control and organization of their functions, some may be a simple as a one neuron layer with an input and an output, and others can mimic complex system such as ANN, which can mimic chromosomal DNA through sizes at cellular level, into artificial organism and simulate reproduction, mutation and population sizes.
1.10 Importance of Data Mining In Biology
Biological data accumulate exponentially in both volume and complexity. The background knowledge increase likely. Automation of knowledge discovery is a part of this accelerating process. This field where application of KDD methodology shows an increasing importance includes annotation of masses of data, structural and functional genomics, protein structure prediction and modeling, analysis biological effects ( function, signaling patterns etc), identification of distantly related proteins, and practical applications (e.g. drug design).
1 Annotation of Masses of Data
The current estimate of the doubling time in both number of entries and sequence base-paers in DNA database is 14-24 months. This is largely because of automated generation of expressed sequence tags (ESTs), which now comprise more than 2/3 of the database entries. Only 5% of estimated 105 human genes have been currently annotated. The components of the gene discovery include: (a) gene identification (b) gene characterization and (c) gene expression. A significant effort has been directed towards building computational tools for gene discovery. These tools include Grail (Uberbacher, et al., 1996)) and Merek gene Index Browse (Eckman, et al., 1998 ).
2. Structural and Functional Genomics
Genomics refers to mapping, sequencing and analysis of complete set of genes and chromosomes in organisms. An initial phase of a genome analysis is construction of high resolution genetic, physical and transcript maps of an organism-structural genomics. The advanced stage comprises the assessment of gene function by making use of the information and reagents provided by structural genomics. According to Hiether & Bogudki, (1997), computational biology will perform a critical and expanding role in this area, whereas structural genomics has been characterized by mining data sets for particularly valuable information. Functional genomics promises to rapidly narrow the gap between sequence and function and to yield new insights into the behaviour of biological systems.
3. Protein Structure, Prediction and Modeling
The structure of a protein can elucidate its function in both general and specific terms, and its evolutionary history, Brenner, et al., (1996). Numerous methods have been developed for protein structure analysis in last two decades. Nevertheless, we still lack the knowledge of the structure for the majority of known proteins.
Secondary and tertiary of structures of only 33% of all sequences SWISS-PROT database are currently available.
· Analysis of Biological Effects
Biological systems are characterized by high degree of complexity and the processes involved are usually multi-step and involve multiple pathways. Sequence databases contain little, if any, higher level knowledge on biological system and processes, but contain voluminous bio-level data. The biological effects need to be studied at higher level of structure. The relevant data are available either as expert knowledge or in the literature source. The high level structure can be encoded in a form of a knowledge or as a model which can then be used to formulate and perform complex quaries promising result in modeling HIV infections were produced by rule-based cellular automation cybermouse which addresses the complexities of the immune system (Siebiry et al., 1993).
· Identification of Distantly Related Proteins
Distant relations between biological sequences provide the main clues for identification and characterization of novel sequences in the database. The approaches include sequence, similarly searches, determination of amino acid motifs, determination of conserved domains and matching sequence patterns. The primary goal is the determination of sequences that display how similarly, but which are significantly related. More sophisticated methods such as Hidden Markov models are gaining popularity
· Practical Applications
Bioinformatics is becoming an important field in drug and vaccine design. Determination of novel compounds for pharmaceutical and agricultural industries includes automated simultaneous screening of very large samples, such as compound collections and combinatorial libraries, termed High throughout screening (HTS). The main challenge in drug discovery research is to identify rapidly novel lead compounds. HTS produces enormous amounts of data which are generally not matched with the ability to analyse these data, creating a bottleneck. KDD and data- mining techniques will keep playing increasingly important role in this domain, Data-mining techniques have been established for determination of peptide candidates for vaccines and immune other apeutic drugs.
KDD process provides a frame works for efficient use of bioinformatics resources in both defining meaningful biological questions and obtaining acceptable answers.
Gominer was developed particularly for biological interpretation of microarray data, one can input a list of under and over expressed genes and a list of all genes on the array, and then calculate enrichment or depletion of categories with genes that have changed expression. Gominer thus facilitates analysis and organization of the results for rapid interpretation of “omic” data. (Weinstein 1998 and Weinstein 2002).
Gominer takes as input two lists of genes: the total set on the array and the subset that the user flags as interesting (for example, altered in expression level), Gominer displays the genes within the frame work of the gene ontology hierarchy, both as a directed graph (DAG) and as the equivalent tree structure. However, each category is annotated assigned to that category plus the number of gene ontology hierarchy.
The most important parameter for purpose of interpretation is the enrichment of interpretation is the enrichment (or depletion) of a category with respect to flagged gene (relative to what would have been expected by chance alone). This parameter will be discussed more extensively and micro mathematically in the section on ‘statistical considerations’.
In Gominer, clicking on a gene of interest in the tree- structures opens a menu that can be used to submit that gene as a query to an external data resource. The number of such links is being expanded rapidly, but currently included are locus link, pubmed, medminer, gene cards, the NCBI’S structure data base and biocarta and KEGG pathway maps as implemented by the NCI Cancer Genome Anatomy Project (CGAP). These external data base provide Gominer with a rich set of resources for bioinformatics integration. For example the links with CGAP and locuslink provide interaction with pathway maps, chromosome visualization, a database of single nucleotide polymorphism (SNP), and the Mammalian gene collection (MGC).
In Gominer, clicking on a category instead of a gene brings up a second visualization, a DAG programmed as a scalable vector graphic (SVG) that can be navigated fluently. Any of its modes can be moused- over to list the flagged genes or clicked to highlight multiple pathways connecting it to the root. Detailed quantitative and statistical results are downloadable in several tab-delimited formats that can be read directly into a text file or a spreadsheet program for further analysis. For example, the spreadsheet data can be sorted by enrichment factor or p- value to focus attention on potentially interesting categories.
1.11 Rationale for the Study
To review the importance of data mining in biology and examine how data mining approaches can be used to understand the mechanisms governing alcohol addiction, a complex disease trait.