Efnisyfirlit

• Cover
• Half Title
• Dedication
• Title Page
• Copyright Page
• Preface
• A Note to the Reader
• Organization of this Book
• Applications and Theory Chapters
• Part 1: Background Basics
• Part 2: Sequence Alignments
• Part 3: Evolutionary Processes
• Part 4: Genome Characteristics
• Part 5: Secondary Structures
• Part 6: Tertiary Structures
• Part 7: Cells and Organisms
• Appendices
• Organization of the Chapters
• Learning Outcomes
• Flow Diagrams
• Mind Maps
• Illustrations
• Further Reading
• List of Symbols
• Glossary
• Garland Science Website
• Artwork
• Additional Material
• List of Reviewers
• Contents In Brief
• Table of Contents
• Part 1 Background Basics
• 1 The Nucleic Acid World
• 1.1 The Structure of DNA and RNA
• DNA is a linear polymer of only four different bases
• Two complementary DNA strands interact by base-pairing to form a double helix
• RNA molecules are mostly single stranded but can also have base-pair structures
• 1.2 DNA, RNA, and Protein: The Central Dogma
• DNA is the information store, but RNA is the messenger
• Messenger RNA is translated into protein according to the genetic code
• Translation involves transfer RNAs and RNA-containing ribosomes
• 1.3 Gene Structure and Control
• RNA polymerase binds to specific sequences that position it and identify where to begin transcription
• The signals initiating transcription in eukaryotes are generally more complex than those in bacteria
• Eukaryotic mRNA transcripts undergo several modifications prior to their use in translation
• The control of translation
• 1.4 The Tree of Life and Evolution
• A brief survey of the basic characteristics of the major forms of life
• Nucleic acid sequences can change as a result of mutation
• Summary
• Further Reading
• General References
• 1.1 The Structure of DNA and RNA
• 1.2 DNA, RNA and Protein: The Central Dogma
• 1.3 Gene Structure and Control
• 1.4 The Tree of Life and Evolution
• Box 1.2
• 2 Protein Structure
• 2.1 Primary and Secondary Structure
• Protein structure can be considered on several different levels
• Amino acids are the building blocks of proteins
• The differing chemical and physical properties of amino acids are due to their side chains
• Amino acids are covalently linked together in the protein chain by peptide bonds
• Secondary structure of proteins is made up of α-helices and β-strands
• Several different types of β-sheet are found in protein structures
• Turns, hairpins, and loops connect helices and strands
• 2.2 Implication for Bioinformatics
• Certain amino acids prefer a particular structural unit
• Evolution has aided sequence analysis
• Visualization and computer manipulation of protein structures
• 2.3 Proteins Fold to Form Compact Structures
• The tertiary structure of a protein is defined by the path of the polypeptide chain
• The stable folded state of a protein represents a state of low energy
• Many proteins are formed of multiple subunits
• Summary
• Further Reading
• 3 Dealing with Databases
• 3.1 The Structure of Databases
• Flat-file databases store data as text files
• Relational databases are widely used for storing biological information
• XML has the flexibility to define bespoke data classifications
• Many other database structures are used for biological data
• Databases can be accessed locally or online and often link to each other
• 3.2 Types of Database
• There’s more to databases than just data
• Primary and derived data
• How we define and connect things is important: Ontologies
• 3.3 Looking for Databases
• Sequence databases
• Microarray databases
• Protein interaction databases
• Structural databases
• 3.4 Data Quality
• Nonredundancy is especially important for some applications of sequence databases
• Automated methods can be used to check for data consistency
• Initial analysis and annotation is usually automated
• Human intervention is often required to produce the highest quality annotation
• The importance of updating databases and entry identifier and version numbers
• Summary
• Further Reading
• 3.1 The Structure of Databases
• 3.2 Types of Database
• How we define and connect things is important: Ontologies
• 3.3 Looking for Databases
• Sequence databases
• Microarray databases
• Protein interaction databases
• Structural databases
• 3.4 Data Quality
• MIAME
• Part 2 Sequence Alignments
• 4 Producing and Analyzing Sequence Alignments
• 4.1 Principles of Sequence Alignment
• Alignment is the task of locating equivalent regions of two or more sequences to maximize their similarity
• Alignment can reveal homology between sequences
• It is easier to detect homology when comparing protein sequences than when comparing nucleic acid sequences
• 4.2 Scoring Alignments
• The quality of an alignment is measured by giving it a quantitative score
• The simplest way of quantifying similarity between two sequences is percentage identity
• The dot-plot gives a visual assessment of similarity based on identity
• Genuine matches do not have to be identical
• There is a minimum percentage identity that can be accepted as significant
• There are many different ways of scoring an alignment
• 4.3 Substitution Matrices
• Substitution matrices are used to assign individual scores to aligned sequence positions
• The PAM substitution matrices use substitution frequencies derived from sets of closely related protein sequences
• The BLOSUM substitution matrices use mutation data from highly conserved local regions of sequence
• The choice of substitution matrix depends on the problem to be solved
• 4.4 Inserting Gaps
• Gaps inserted in a sequence to maximize similarity with another require a scoring penalty
• Dynamic programming algorithms can determine the optimal introduction of gaps
• 4.5 Types of Alignment
• Different kinds of alignments are useful in different circumstances
• Multiple sequence alignments enable the simultaneous comparison of a set of similar sequences
• Multiple alignments can be constructed by several different techniques
• Multiple alignments can improve the accuracy of alignment for sequences of low similarity
• ClustalW can make global multiple alignments of both DNA and protein sequences
• Multiple alignments can be made by combining a series of local alignments
• Alignment can be improved by incorporating additional information
• 4.6 Searching Databases
• Fast yet accurate search algorithms have been developed
• FASTA is a fast database-search method based on matching short identical segments
• BLAST is based on finding very similar short segments
• Different versions of BLAST and FASTA are used for different problems
• PSI-BLAST enables profile-based database searches
• Ssearch is a rigorous alignment method
• 4.7 Searching with Nucleic Acid or Protein Sequences
• DNA or RNA sequences can be used either directly or after translation
• The quality of a database match has to be tested to ensure that it could not have arisen by chance
• Choosing an appropriate E-value threshold helps to limit a database search
• Low-complexity regions can complicate homology searches
• Different databases can be used to solve particular problems
• 4.8 Protein Sequence Motifs or Patterns
• Creation of pattern databases requires expert knowledge
• The BLOCKS database contains automatically compiled short blocks of conserved multiply aligned protein sequences
• 4.9 Searching Using Motifs and Patterns
• The PROSITE database can be searched for protein motifs and patterns
• The pattern-based program PHI-BLAST searches for both homology and matching motifs
• Patterns can be generated from multiple sequences using PRATT
• The PRINTS database consists of fingerprints representing sets of conserved motifs that describe a protein family
• The Pfam database defines profiles of protein families
• 4.10 Patterns and Protein Function
• Searches can be made for particular functional sites in proteins
• Sequence comparison is not the only way of analyzing protein sequences
• Summary
• Further Reading
• 4.1 Principles of Sequence Alignment
• 4.2 Scoring Alignments
• Twilight zone and midnight zone
• 4.3 Substitution Matrices
• 4.4 Inserting Gaps
• 4.5 Types of Alignment
• ClustalW
• DIALIGN
• 4.6 Searching Databases
• BLAST
• FASTA
• PSI-BLAST
• 4.8 Protein Sequence Motifs or Patterns
• MEME
• 4.9 Searching Using Motifs and Patterns
• MOTIF
• PRATT
• 4.10 Patterns and Protein Function
• HCA
• 5 Pairwise Sequence Alignment and Database Searching
• 5.1 Substitution Matrices and Scoring
• Alignment scores attempt to measure the likelihood of a common evolutionary ancestor
• The PAM (MDM) substitution scoring matrices were designed to trace the evolutionary origins of proteins
• The BLOSUM matrices were designed to find conserved regions of proteins
• Scoring matrices for nucleotide sequence alignment can be derived in similar ways
• The substitution scoring matrix used must be appropriate to the specific alignment problem
• Gaps are scored in a much more heuristic way than substitutions
• 5.2 Dynamic Programming Algorithms
• Optimal global alignments are produced using efficient variations of the Needleman-Wunsch algorithm
• Local and suboptimal alignments can be produced by making small modifications to the dynamic programming algorithm
• Time can be saved with a loss of rigor by not calculating the whole matrix
• 5.3 Indexing Techniques and Algorithmic Approximations
• Suffix trees locate the positions of repeats and unique sequences
• Hashing is an indexing technique that lists the starting positions of all k-tuples
• The FASTA algorithm uses hashing and chaining for fast database searching
• The BLAST algorithm makes use of finite-state automata
• Comparing a nucleotide sequence directly with a protein sequence requires special modifications to the BLAST and FASTA algorithms
• 5.4 Alignment Score Significance
• The statistics of gapped local alignments can be approximated by the same theory
• 5.5 Aligning Complete Genome Sequences
• Indexing and scanning whole genome sequences efficiently is crucial for the sequence alignment of higher organisms
• The complex evolutionary relationships between the genomes of even closely related organisms require novel alignment algorithms
• Summary
• Further Reading
• 5.1 Substitution Matrices and Scoring
• The PAM (MDM) substitution scoring matrices were designed to trace the evolutionary origins of proteins
• The BLOSUM matrices were designed to find conserved regions of proteins
• Scoring matrices for nucleotide sequence alignment can be derived in similar ways
• The substitution scoring matrix used must be appropriate to the specific alignment problem
• Gaps are scored in a much more heuristic way than substitutions
• 5.2 Dynamic Programming Algorithms
• Optimal global alignments are produced using efficient variations of the Needleman-Wunsch algorithm
• Local and suboptimal alignments can be produced by making small modifications to the dynamic programming algorithm
• Time can be saved with a loss of rigor by not calculating the whole matrix
• 5.3 Indexing Techniques and Algorithmic Approximations
• Suffix trees locate the positions of repeats and unique sequences; Hashing is an indexing technique that lists the starting positions of all k-tuples
• The FASTA algorithm uses hashing and chaining for fast database searching; The BLAST algorithm makes use of finite-state automata
• Box 5.2: Sometimes things just aren’t complex enough
• 5.4 Alignment Score Significance
• 5.5 Alignments Involving Complete Genome Sequences
• 6 Patterns, Profiles, and Multiple Alignments
• 6.1 Profiles and Sequence Logos
• Position-specific scoring matrices are an extension of substitution scoring matrices
• Methods for overcoming a lack of data in deriving the values for a PSSM
• PSI-BLAST is a sequence database searching program
• Representing a profile as a logo
• 6.2 Profile Hidden Markov Models
• The basic structure of HMMs used in sequence alignment to profiles
• Estimating HMM parameters using aligned sequences
• Scoring a sequence against a profile HMM: The most probable path and the sum over all paths
• Estimating HMM parameters using unaligned sequences
• 6.3 Aligning Profiles
• Comparing two PSSMs by alignment
• Aligning profile HMMs
• 6.4 Multiple Sequence Alignments by Gradual Sequence Addition
• The order in which sequences are added is chosen based on the estimated likelihood of incorporating errors in the alignment
• Many different scoring schemes have been used in constructing multiple alignments
• The multiple alignment is built using the guide tree and profile methods and may be further refined
• 6.5 Other Ways of Obtaining Multiple Alignments
• The multiple sequence alignment program DIALIGN aligns ungapped blocks
• The SAGA method of multiple alignment uses a genetic algorithm
• 6.6 Sequence Pattern Discovery
• Discovering patterns in a multiple alignment: eMOTIF and AACC
• Probabilistic searching for common patterns in sequences: Gibbs and MEME
• Searching for more general sequence patterns
• Summary
• Further Reading
• 6.1 Profiles and Sequence Logos
• 6.2 Profile Hidden Markov Models
• 6.3 Aligning Profiles
• 6.4 Multiple Sequence Alignments by Gradual Sequence Additions
• 6.5 Other Ways of Obtaining Multiple Alignments
• 6.6 Sequence Pattern Discovery
• Part 3 Evolutionary Processes
• 7 Recovering Evolutionary History
• 7.1 The Structure and Interpretation of Phylogenetic Trees
• Phylogenetic trees reconstruct evolutionary relationships
• Tree topology can be described in several ways
• Consensus and condensed trees report the results of comparing tree topologies
• 7.2 Molecular Evolution and its Consequences
• Most related sequences have many positions that have mutated several times
• The rate of accepted mutation is usually not the same for all types of base substitution
• Different codon positions have different mutation rates
• Only orthologous genes should be used to construct species phylogenetic trees
• Major changes affecting large regions of the genome are surprisingly common
• 7.3 Phylogenetic Tree Reconstruction
• Small ribosomal subunit rRNA sequences are well suited to reconstructing the evolution of species
• The choice of the method for tree reconstruction depends to some extent on the size and quality of the dataset
• A model of evolution must be chosen to use with the method
• All phylogenetic analyses must start with an accurate multiple alignment
• Phylogenetic analyses of a small dataset of 16S RNA sequence data
• Building a gene tree for a family of enzymes can help to identify how enzymatic functions evolved
• Summary
• Further Reading
• General
• 7.1 The Structure and Interpretation of Phylogenetic Trees
• Phylogenetic trees reconstruct evolutionary relationships
• Tree topology can be described in several ways
• Consensus and condensed trees report the results of comparing tree topologies
• 7.2 Molecular Evolution and its Consequences
• The rate of accepted mutation is usually not the same for all types of base substitution
• Different codon positions have different mutation rates
• Only orthologous genes should be used to construct species phylogenetic trees
• Major changes affecting large regions of the genome are surprisingly common
• 7.3 Phylogenetic Tree Reconstruction
• Small ribosomal subunit rRNA sequences are well suited to reconstructing the evolution of species
• The choice of the method for tree reconstruction depends to some extent on the size and quality of the dataset
• A model of evolution must be chosen to use with the method
• Phylogenetic analyses of a small dataset of 16S RNA sequence data
• Building a gene tree for a family of enzymes can help to identify how enzymatic functions evolved
• 8 Building Phylogenetic Trees
• 8.1 Evolutionary Models and the Calculation of Evolutionary Distance
• A simple but inaccurate measure of evolutionary distance is the p-distance
• The Poisson distance correction takes account of multiple mutations at the same site
• The Gamma distance correction takes account of mutation rate variation at different sequence positions
• The Jukes-Cantor model reproduces some basic features of the evolution of nucleotide sequences
• More complex models distinguish between the relative frequencies of different types of mutation
• There is a nucleotide bias in DNA sequences
• Models of protein-sequence evolution are closely related to the substitution matrices used for sequence alignment
• 8.2 Generating Single Phylogenetic Trees
• Clustering methods produce a phylogenetic tree based on evolutionary distances
• The UPGMA method assumes a constant molecular clock and produces an ultrametric tree
• The Fitch-Margoliash method produces an unrooted additive tree
• The neighbor-joining method is related to the concept of minimum evolution
• Stepwise addition and star-decomposition methods are usually used to generate starting trees for further exploration, not the final tree
• 8.3 Generating Multiple Tree Topologies
• The branch-and-bound method greatly improves the efficiency of exploring tree topology
• Optimization of tree topology can be achieved by making a series of small changes to an existing tree
• Finding the root gives a phylogenetic tree a direction in time
• 8.4 Evaluating Tree Topologies
• Functions based on evolutionary distances can be used to evaluate trees
• Unweighted parsimony methods look for the trees with the smallest number of mutations
• Mutations can be weighted in different ways in the parsimony method
• Trees can be evaluated using the maximum likelihood method
• The quartet-puzzling method also involves maximum likelihood in the standard implementation
• Bayesian methods can also be used to reconstruct phylogenetic trees
• 8.5 Assessing the Reliability of Tree Features and Comparing Trees
• The long-branch attraction problem can arise even with perfect data and methodology
• Tree topology can be tested by examining the interior branches
• Tests have been proposed for comparing two or more alternative trees
• Summary
• Further Reading
• 8.1 Evolutionary Models and the Calculation of Evolutionary Distance
• The Gamma distance correction takes account of mutation rate variation at different sequence positions
• More complex models distinguish between the relative frequencies of different types of mutation
• There is a nucleotide bias in DNA sequences
• Models of protein-sequence evolution are closely related to the substitution matrices used for sequence alignment
• 8.2 Generating Single Phylogenetic Trees
• The UPGMA method assumes a constant molecular clock and produces an ultrametric tree
• The Fitch-Margoliash method produces an unrooted additive tree
• The neighbor-joining method is related to the concept of minimum evolution
• 8.3 Generating Multiple Tree Topologies
• Optimization of tree topology can be achieved by making a series of small changes to an existing tree
• Finding the root gives a phylogenetic tree a direction in time
• 8.4 Evaluating Tree Topologies
• Functions based on evolutionary distances can be used to evaluate trees
• Trees can be evaluated using the maximum likelihood method
• The quartet-puzzling method also involves maximum likelihood in the standard implementation
• Bayesian methods can also be used to reconstruct phylogenetic trees
• 8.5 Assessing the Reliability of Tree Features and Comparing Trees
• The long-branch attraction problem can arise even with perfect data and methodology
• Tree topology can be tested by examining the interior branches
• Tests have been proposed for comparing two or more alternative trees
• Part 4 Genome Characteristics
• 9 Revealing Genome Features
• 9.1 Preliminary Examination of Genome Sequence
• Whole genome sequences can be split up to simplify gene searches
• Structural RNA genes and repeat sequences can be excluded from further analysis
• Homology can be used to identify genes in both prokaryotic and eukaryotic genomes
• 9.2 Gene Prediction in Prokaryotic Genomes
• 9.3 Gene Prediction in Eukaryotic Genomes
• Programs for predicting exons and introns use a variety of approaches
• Gene predictions must preserve the correct reading frame
• Some programs search for exons using only the query sequence and a model for exons
• Some programs search for genes using only the query sequence and a gene model
• Genes can be predicted using a gene model and sequence similarity
• Genomes of related organisms can be used to improve gene prediction
• 9.4 Splice Site Detection
• Splice sites can be detected independently by specialized programs
• 9.5 Prediction of Promoter Regions
• Prokaryotic promoter regions contain relatively well-defined motifs
• Eukaryotic promoter regions are typically more complex than prokaryotic promoters
• A variety of promoter-prediction methods are available online
• Promoter prediction results are not very clear-cut
• 9.6 Confirming Predictions
• There are various methods for calculating the accuracy of gene-prediction programs
• Translating predicted exons can confirm the correctness of the prediction
• Constructing the protein and identifying homologs
• 9.7 Genome Annotation
• Genome annotation is the final step in genome analysis
• Gene ontology provides a standard vocabulary for gene annotation
• 9.8 Large Genome Comparisons
• Summary
• Further Reading
• 9.1 Preliminary Examination of Genome Sequence
• 9.2 Gene Prediction in Prokaryotic Genomes
• 9.3 Gene Prediction in Eukaryotic Genomes
• 9.4 Splice Site Detection
• 9.5 Prediction of Promoter Regions
• 9.6 Confirming Predictions
• 9.7 Genome Annotation
• 9.8 Large Genome Comparisons
• Box 9.5
• 10 Gene Detection and Genome Annotation
• 10.1 Detection of Functional RNA Molecules Using Decision Trees
• Detection of tRNA genes using the tRNAscan algorithm
• Detection of tRNA genes in eukaryotic genomes
• 10.2 Features Useful for Gene Detection in Prokaryotes
• 10.3 Algorithms for Gene Detection in Prokaryotes
• GeneMark uses inhomogeneous Markov chains and dicodon statistics
• GLIMMER uses interpolated Markov models of coding potential
• Orpheus uses homology, codon statistics, and ribosome-binding sites
• GeneMark.hmm uses explicit state duration hidden Markov models
• EcoParse is an HMM gene model
• 10.4 Features Used in Eukaryotic Gene Detection
• Differences between prokaryotic and eukaryotic genes
• Introns, exons, and splice sites
• Promoter sequences and binding sites for transcription factors
• 10.5 Predicting Eukaryotic Gene Signals
• Detection of core promoter binding signals is a key element of some eukaryotic gene-prediction methods
• A set of models has been designed to locate the site of core promoter sequence signals
• Predicting promoter regions from general sequence properties can reduce the numbers of false-positive results
• Predicting eukaryotic transcription and translation start sites
• Translation and transcription stop signals complete the gene definition
• 10.6 Predicting Exons and Introns
• Exons can be identified using general sequence properties
• Splice-site prediction
• Splice sites can be predicted by sequence patterns combined with base statistics
• GenScan uses a combination of weight matrices and decision trees to locate splice sites
• GeneSplicer predicts splice sites using first-order Markov chains
• NetPlantGene combines neural networks with intron and exon predictions to predict splice sites
• Other splicing features may yet be exploited for splice-site prediction
• Specific methods exist to identify initial and terminal exons
• Exons can be defined by searching databases for homologous regions
• 10.7 Complete Eukaryotic Gene Models
• 10.8 Beyond the Prediction of Individual Genes
• Functional annotation
• Comparison of related genomes can help resolve uncertain predictions
• Evaluation and reevaluation of gene-detection methods
• Summary
• Further Reading
• 10.1 Detection of Functional RNA Molecules Using Decision Trees
• tRNA detection
• Detection of other RNA genes
• 10.2 Features Useful for Gene Detection in Prokaryotes
• Identifying protein-coding regions using base statistics
• 10.3 Algorithms for Gene Detection in Prokaryotes
• GeneMark, GeneMark.hmm, and further developments
• Glimmer
• Orpheus
• EcoParse
• Prokaryotic genomes
• Markov models
• 10.4 Features Used in Eukaryotic Gene Detection
• 10.4 Preliminary analysis for human genes
• 10.5 Predicting Eukaryotic Gene Signals
• Initial analysis of core promoter sequences
• Algorithms of core promoter detection
• Promoter recognition
• 10.6 Predicting Exons and Introns
• 10.7 Complete Eukaryotic Gene Models
• 10.8 Beyond the Prediction of Individual Genes
• Detailed reexamination of the annotation of a complete genome
• Large-scale changes in chromosomes
• Box 10.1 Measures of gene prediction accuracy at the nucleotide level
• Box 10.2 Sequencing many genomes at once
• Box 10.3 Measures of gene prediction accuracy at the exon level
• Part 5 Secondary Structures
• 11 Obtaining Secondary Structure from Sequence
• 11.1 Types of Prediction Methods
• Statistical methods are based on rules that give the probability that a residue will form part of a particular secondary structure
• Nearest-neighbor methods are statistical methods that incorporate additional information about protein structure
• Machine-learning approaches to secondary structure prediction mainly make use of neural networks and HMM methods
• 11.2 Training and Test Databases
• There are several ways to define protein secondary structures
• 11.3 Assessing the Accuracy of Prediction Programs
• Q3 measures the accuracy of individual residue assignments
• Secondary structure predictions should not be expected to reach 100% residue accuracy
• The Sov value measures the prediction accuracy for whole elements
• CAFASP/CASP: Unbiased and readily available protein prediction assessments
• 11.4 Statistical and Knowledge-Based Methods
• The GOR method uses an information theory approach
• The program Zpred includes multiple alignment of homologous sequences and residue conservation information
• There is an overall increase in prediction accuracy using multiple sequence information
• The nearest-neighbor method: The use of multiple nonhomologous sequences
• Predator is a combined statistical and knowledge-based program that includes the nearest-neighbor approach
• 11.5 Neural Network Methods of Secondary Structure Prediction
• Assessing the reliability of neural net predictions
• Several examples of Web-based neural network secondary structure prediction programs
• PROF: Protein forecasting
• Psipred
• Jnet: Using several alternative representations of the sequence alignment
• 11.6 Some Secondary Structures Require Specialized Prediction Methods
• Transmembrane proteins
• Quantifying the preference for a membrane environment
• 11.7 Prediction of Transmembrane Protein Structure
• Multi-helix membrane proteins
• A selection of prediction programs to predict transmembrane helices
• Statistical methods
• Knowledge-based prediction
• Evolutionary information from protein families improves the prediction
• Neural nets in transmembrane prediction
• Predicting transmembrane helices with hidden Markov models
• Comparing the results: What to choose
• What happens if a non-transmembrane protein is submitted to transmembrane prediction programs
• Prediction of transmembrane structure containing β-strands
• 11.8 Coiled-Coil Structures
• The COILS prediction program
• PAIRCOIL and MULTICOIL are an extension of the COILS algorithm
• Zipping the leucine zipper: A specialized coiled coil
• 11.9 RNA Secondary Structure Prediction
• Summary
• Further Reading
• 11.1 Types of Prediction Methods
• 11.2 Training and Test Databases
• PDB
• STRIDE
• DSSP
• DEFINE
• 11.3 Assessing the Accuracy of Prediction Programs
• 11.4 Statistical and Knowledge-Based Methods
• 11.5 Neural Network Methods of Secondary Structure Prediction
• 11.7 Prediction of Transmembrane Protein Structure
• 11.8 Coiled-Coil Structures
• 11.9 RNA Secondary Structure Prediction
• Box 11.1
• Box 11.3
• 12 Predicting Secondary Structures
• 12.1 Defining Secondary Structure and Prediction Accuracy
• The definitions used for automatic protein secondary structure assignment do not give identical results
• There are several different measures of the accuracy of secondary structure prediction
• 12.2 Secondary Structure Prediction Based on Residue Propensities
• Each structural state has an amino acid preference which can be assigned as a residue propensity
• The simplest prediction methods are based on the average residue propensity over a sequence window
• Residue propensities are modulated by nearby sequence
• Predictions can be significantly improved by including information from homologous sequences
• 12.3 The Nearest-Neighbor Methods are Based on Sequence Segment Similarity
• Short segments of similar sequence are found to have similar structure
• Several sequence similarity measures have been used to identify nearest-neighbor segments
• A weighted average of the nearest-neighbor segment structures is used to make the prediction
• A nearest-neighbor method has been developed to predict regions with a high potential to misfold
• 12.4 Neural Networks Have Been Employed Successfully for Secondary Structure Prediction
• Layered feed-forward neural networks can transform a sequence into a structural prediction
• Inclusion of information on homologous sequences improves neural network accuracy
• More complex neural nets have been applied to predict secondary and other structural features
• 12.5 Hidden Markov Models Have Been Applied to Structure Prediction
• HMM methods have been found especially effective for transmembrane proteins
• Nonmembrane protein secondary structures can also be successfully predicted with HMMs
• 12.6 General Data Classification Techniques can Predict Structural Features
• Support vector machines have been successfully used for protein structure prediction
• Discriminants, SOMs, and other methods have also been used
• Summary
• Further Reading
• 12.1 Defining Secondary Structure and Prediction Accuracy
• DSSP
• PALSSE
• β-Spider
• Limits of prediction accuracy
• TM properties and accuracy measures
• PSIPRED (Q3 and Sov variation)
• Matthews correlation coefficient
• Sov
• SS assignment comparison
• Length distributions (non-TM)
• 12.2 Secondary Structure Prediction Based on Residue Propensities
• PDB_SELECT
• In-depth analysis of unbiased structural datasets
• Hydrophobicity scales
• Chou-Fasman
• COILS
• MEMSAT
• Local and nonlocal effects
• β-turn propensities
• β-turn propensities and use of PSSMs
• AAindex
• GOR theory
• GOR I
• GOR II
• GOR III
• GOR IV
• GOR V (includes other sequences)
• Using consecutive pair of structural states
• Zpred
• Treatment of gaps in multiple alignments during prediction
• 12.3 The Nearest-neighbor Methods are Based on Sequence Segment Similarity
• NNSSP
• SSPAL
• I-sites
• SIMPA96
• Correlation between amino acid composition and protein structural class
• HβP
• 12.4 Neural Networks Have Been Employed Successfully for Secondary Structure Prediction
• PHDsec
• PHDpsi
• PSIPRED
• Back-propagation learning
• BTPRED
• Betaturns
• DESTRUCT
• SSpro
• 12.5 Hidden Markov Models Have Been Applied to Structure Prediction
• HMMTOP
• TMHMM
• Phobius
• PROftmb
• PRED-TMBB
• YASPIN
• MARCOIL
• 12.6 General Data Classification Techniques can Predict Structural Features
• DSC
• FoldIndex
• GPI-SOM
• PSIMLR
• Part 6 Tertiary Structures
• 13 Modeling Protein Structure
• 13.1 Potential Energy Functions and Force Fields
• The conformation of a protein can be visualized in terms of a potential energy surface
• Conformational energies can be described by simple mathematical functions
• Similar force fields can be used to represent conformational energies in the presence of averaged environments
• Potential energy functions can be used to assess a modeled structure
• Energy minimization can be used to refine a modeled structure and identify local energy minima
• Molecular dynamics and simulated annealing are used to find global energy minima
• 13.2 Obtaining a Structure by Threading
• The prediction of protein folds in the absence of known structural homologs
• Libraries or databases of nonredundant protein folds are used in threading
• Two distinct types of scoring schemes have been used in threading methods
• Dynamic programming methods can identify optimal alignments of target sequences and structural folds
• Several methods are available to assess the confidence to be put on the fold prediction
• The C2-like domain from the Dictyostelia: A practical example of threading
• 13.3 Principles of Homology Modeling
• Closely related target and template sequences give better models
• Significant sequence identity depends on the length of the sequence
• Homology modeling has been automated to deal with the numbers of sequences that can now be modeled
• Model building is based on a number of assumptions
• 13.4 Steps in Homology Modeling
• Structural homologs to the target protein are found in the PDB
• Accurate alignment of target and template sequences is essential for successful modeling
• The structurally conserved regions of a protein are modeled first
• The modeled core is checked for misfits before proceeding to the next stage
• Sequence realignment and remodeling may improve the structure
• Insertions and deletions are usually modeled as loops
• Nonidentical amino acid side chains are modeled mainly by using rotamer libraries
• Energy minimization is used to relieve structural errors
• Molecular dynamics can be used to explore possible conformations for mobile loops
• Models need to be checked for accuracy
• How far can homology models be trusted?
• 13.5 Automated Homology Modeling
• The program MODELLER models by satisfying protein structure constraints
• COMPOSER uses fragment-based modeling to automatically generate a model
• Automated methods available on the Web for comparative modeling
• Assessment of structure prediction
• 13.6 Homology Modeling of PI3 Kinase p110a
• Swiss-Pdb Viewer can be used for manual or semi-manual modeling
• Alignment, core modeling, and side-chain modeling are carried out all in one
• The loops are modeled from a database of possible structures
• Energy minimization and quality inspection can be carried out within Swiss-Pdb Viewer
• MolIDE is a downloadable semi-automatic modeling package
• Automated modeling on the Web illustrated with p110α kinase
• Modeling a functionally related but sequentially dissimilar protein: mTOR
• Generating a multidomain three-dimensional structure from sequence
• Summary
• Further Reading
• Modeling: General
• 13.1 Potential Energy Functions and Force Fields
• Ab initio modeling
• 13.2 Obtaining a Structure by Threading
• 123D+
• GenTHREADER
• 3D-PSSM
• FUGUE
• LIBRA
• LOOPP
• LIBELLULA
• SCOP
• 13.3 Automated Homology Modeling
• MolIDE
• SCWRL
• PSIPRED
• LOOPY
• MODELLER
• Assessing models
• MolProbity
• PROCHECK
• CASP
• Antibody and HIV examples
• 14 Analyzing Structure-Function Relationships
• 14.1 Functional Conservation
• Functional regions are usually structurally conserved
• Similar biochemical function can be found in proteins with different folds
• Fold libraries identify structurally similar proteins regardless of function
• 14.2 Structure Comparison Methods
• Finding domains in proteins aids structure comparison
• Structural comparisons can reveal conserved functional elements not discernible from a sequence comparison
• The CE method builds up a structural alignment from pairs of aligned protein segments
• The Vector Alignment Search Tool (VAST) aligns secondary structural elements
• DALI identifies structure superposition without maintaining segment order
• FATCAT introduces rotations between rigid segments
• 14.3 Finding Binding Sites
• Highly conserved, strongly charged, or hydrophobic surface areas may indicate interaction sites
• Searching for protein-protein interactions using surface properties
• Surface calculations highlight clefts or holes in a protein that may serve as binding sites
• Looking at residue conservation can identify binding sites
• 14.4 Docking Methods and Programs
• Simple docking procedures can be used when the structure of a homologous protein bound to a ligand analog is known
• Specialized docking programs will automatically dock a ligand to a structure
• Scoring functions are used to identify the most likely docked ligand
• The DOCK program is a semirigid-body method that analyzes shape and chemical complementarity of ligand and binding site
• Fragment docking identifies potential substrates by predicting types of atoms and functional groups in the binding area
• GOLD is a flexible docking program, which utilizes a genetic algorithm
• The water molecules in binding sites should also be considered
• Summary
• Further Reading
• General
• 14.1 Functional Conservation
• Structure-function relationships
• Fold recognition programs
• Domain identification
• Viewing programs
• 14.3 Finding Binding Sites
• Evolutionary trace methods
• 14.4 Docking Methods and Programs
• Part 7 Cells and Organisms
• 15 Proteome and Gene Expression Analysis
• 15.1 Analysis of Large-scale Gene Expression
• The expression of large numbers of different genes can be measured simultaneously by DNA microarrays
• Gene expression microarrays are mainly used to detect differences in gene expression in different conditions
• Serial analysis of gene expression (SAGE) is also used to study global patterns of gene expression
• Digital differential display uses bioinformatics and statistics to detect differential gene expression in different tissues
• Facilitating the integration of data from different places and experiments
• The simplest method of analyzing gene expression microarray data is hierarchical cluster analysis
• Techniques based on self-organizing maps can be used for analyzing microarray data
• Self-organizing tree algorithms (SOTAs) cluster from the top down by successive subdivision of clusters
• Clustered gene expression data can be used as a tool for further research
• 15.2 Analysis of Large-scale Protein Expression
• Two-dimensional gel electrophoresis is a method for separating the individual proteins in a cell
• Measuring the expression levels shown in 2D gels
• Differences in protein expression levels between different samples can be detected by 2D gels
• Clustering methods are used to identify protein spots with similar expression patterns
• Principal component analysis (PCA) is an alternative to clustering for analyzing microarray and 2D gel data
• The changes in a set of protein spots can be tracked over a number of different samples
• Databases and online tools are available to aid the interpretation of 2D gel data
• Protein microarrays allow the simultaneous detection of the presence or activity of large numbers of different proteins
• Mass spectrometry can be used to identify the proteins separated and purified by 2D gel electrophoresis or other means
• Protein-identification programs for mass spectrometry are freely available on the Web
• Mass spectrometry can be used to measure protein concentration
• Summary
• Further Reading
• 15.1 Analysis of Large-scale Gene Expression
• Functional genomics
• Microarray quality control focus
• MIAME
• 15.2 Analysis of Large-scale Protein Expression
• Proteomics
• Protein microarrays
• MASCOT
• ProteinProspector
• Databases and software
• 16 Clustering Methods and Statistics
• 16.1 Expression Data Require Preparation Prior to Analysis
• Data normalization is designed to remove systematic experimental errors
• Expression levels are often analyzed as ratios and are usually transformed by taking logarithms
• Sometimes further normalization is useful after the data transformation
• Principal component analysis is a method for combining the properties of an object
• 16.2 Cluster Analysis Requires Distances to be Defined Between all Data Points
• Euclidean distance is the measure used in everyday life
• The Pearson correlation coefficient measures distance in terms of the shape of the expression response
• The Mahalanobis distance takes account of the variation and correlation of expression responses
• 16.3 Clustering Methods Identify Similar and Distinct Expression Patterns
• Hierarchical clustering produces a related set of alternative partitions of the data
• k-means clustering groups data into several clusters but does not determine a relationship between clusters
• Self-organizing maps (SOMs) use neural network methods to cluster data into a predetermined number of clusters
• Evolutionary clustering algorithms use selection, recombination, and mutation to find the best possible solution to a problem
• The self-organizing tree algorithm (SOTA) determines the number of clusters required
• Biclustering identifies a subset of similar expression level patterns occurring in a subset of the samples
• The validity of clusters is determined by independent methods
• 16.4 Statistical Analysis can Quantify the Significance of Observed Differential Expression
• T-tests can be used to estimate the significance of the difference between two expression levels
• Nonparametric tests are used to avoid making assumptions about the data sampling
• Multiple testing of differential expression requires special techniques to control error rates
• 16.5 Gene and Protein Expression Data Can be Used to Classify Samples
• Many alternative methods have been proposed that can classify samples
• Support vector machines are another form of supervised learning algorithm that can produce classifiers
• Summary
• Further Reading
• Monograph
• 16.1 Expression Data Require Preparation Prior to Analysis
• Variance-stabilizing transformation
• Lowess normalization
• General review of normalization and transformation of data
• Data used to generate Figures 16.2 to 16.4
• Principal component analysis
• 16.2 Cluster Analysis Requires Distances to be Defined Between all Data Points
• Distance definitions
• 16.3 Clustering Methods Identify Similar and Distinct Expression Patterns
• Self-organizing maps (SOMs)
• Clustering using genetic algorithms
• SOTA
• Biclustering techniques
• Validation of partitions produced by clustering methods
• 16.4 Statistical Analysis can Quantify the Significance of Observed Differential Expression
• Bayesian estimation of variance
• Nonparametric tests
• Multiple hypothesis testing
• 16.5 Gene and Protein Expression Data Can be Used to Classify Samples
• Support vector machines
• 17 Systems Biology
• 17.1 What is a System?
• A system is more than the sum of its parts
• A biological system is a living network
• Databases are useful starting points in constructing a network
• To construct a model more information is needed than a network
• There are three possible approaches to constructing a model
• Kinetic models are not the only way in systems biology
• 17.2 Structure of the Model
• Control circuits are an essential part of any biological system
• The interactions in networks can be represented as simple differential equations
• 17.3 Robustness of Biological Systems
• Robustness is a distinct feature of complexity in biology
• Modularity plays an important part in robustness
• Redundancy in the system can provide robustness
• Living systems can switch from one state to another by means of bistable switches
• 17.4 Storing and Running System Models
• Specialized programs make simulating systems easier
• Standardized systems descriptions aid their storage and reuse
• Summary
• Further Reading
• 17.1 What is a System?
• Databases
• 17.2 Structure of the Model
• Topological modeling
• 17.3 Robustness of Biological Systems
• Molecular systems modeling
• Appendix A: probability, Information, and Bayesian Analysis
• Probability Theory, Entropy, and Information
• Mutually exclusive events
• Occurrence of two events
• Occurrence of two random variables
• Bayesian Analysis
• Bayes’ theorem
• Inference of parameter values
• Further Reading
• Appendix B: Molecular Energy Functions
• Force Fields for Calculating Intra- and Intermolecular Interaction Energies
• Bonding terms
• Nonbonding terms
• Potentials used in Threading
• Potentials of mean force
• Potential terms relating to solvent effects
• Further Reading
• Appendix C: Function Optimization
• Full Search Methods
• Dynamic programming and branch-and-bound
• Local Optimization
• The downhill simplex method
• The steepest descent method
• The conjugate gradient method
• Methods using second derivatives
• Thermodynamic Simulation and Global Optimization
• Monte Carlo and genetic algorithms
• Molecular dynamics
• Simulated annealing
• Summary
• Further Reading
• List of Symbols
• Notation concepts
• Glossary
• Index