An Overview of Bioinformatics

1
An Overview of Bioinformatics

• Yuh-Jyh Hu
• CIS, NCTU

2
Overview

• Genomics, Bioinformatics and Medicine

Molecular Diagnosis
Genomics
Identify Drug Targets
Molecular Epidemiology
Genetic Therapy
Rational Drug Design
Bioinformatics
Information Theory
Graph Theory
Artificial Intelligence
Robotics
Machine Learning
Databases
Statistics
Algorithm
3

• Introduction to Bioinformatics
• Bioinformatics a strategic discipline at the
  frontier between biology and computer science
• Loosely defined at the intersection of molecular
  and computational biology
• Primarily contributed by the academic user
  community
• Driving force the advent of new, efficient
  experimental techniques, especially in DNA
  sequencing
• Major goals understanding of life and evolution,
  and discoveries of new drugs and therapies.

4

• Significant advances
• Collecting and managing data
• Databases of various types
• Nucleotide sequences
• Protein sequences
• Structures
• Gene expression
• Integrated data retrieval
• Entrez at NCBI
• Data Analysis
• Sequence
• Structure
• Expression

5
(No Transcript)
6
(No Transcript)
7

• Entrez
• Developed at NCBI, freely available and allow for
  integrated access to PubMed records, nucleotide
  and protein sequence data and 3D structure
  information
• http//www.ncbi.nlm.nih.gov/Entrez

8
(No Transcript)
9
Central Dogma of Molecular Biology
transcription
translation
replication
DNA
RNA
Protein
reverse transcription

• Molecules
• Structure
• Function
• Processes
• Mechanism
• Regulation

10
What Information to organize?

• Molecular Biology as Information Science
• Central Dogma of Molecular Biology Central
  Paradigm for Bioinformatics
• DNA Genomic Sequence
• ? ?
• RNA mRNA (gene expression level)
• ? ?
• Protein Protein Sequence
• ? ?
• Phenotype Protein Function
• ?
• Phenotype
• Biological Process Process large amount and
  different
• types of information
• Performed or facilitated by proteins,
  apply various informatics techniques
• executing instructions encoded in DNA

11
Organize Information

• Redundancy and Multiplicity
• Different sequences have the same/similar
  structure
• Organism has many similar genes
• Single may have multiple functions
• How to find the similarities?
• Modular parts
• shared parts (bolts, washers)
• unique parts (steering wheel)
• Vast growth in data but limited increase of
  fundamentally new parts
• More new protein structures, but not many new
  protein families
• Simplification by grouping conserved elements
  (e.g. sequences, structures) into modular parts

12

• Examining and analyzing data
• Gene finding
• ML techniques have been applied to almost all
  steps in computational gene finding, including
  the assignment of translation start and stop,
  quantification of reading frames, gene modeling,
  etc.
• Promoter recognition transcription initiation
  and termination
• Transcription initiation the first step in gene
  expression
• RNA polymerase recognizes and binds to certain
  sequences called promoters.
• Why difficult?
• Large variable distance between DNA signals
  recognized by RNA polymerase
• Many other factors involved in expression
  regulation

13

• Gene Expression
• Clustering group the genes with similar
  expression behavior together
• Find target genes correlated with diseases
• Construct genetic network understanding gene
  interactions
• Motif prediction patterns showing regularity
• Sequence motifs DNA, proteins
• Structural motifs RNA, proteins
• Protein structure prediction
• Classification
• Protein family
• DNA-protein interaction

14

• Information Retrieval from Biological databases
• Exponential growth of databases
• GeneBank is an annotated collection of all
  publicly available DNA sequences. It contains
  over 1.6 million sequence records covering over 1
  billion nucleotides.
• Many efforts have been driven into making such
  data accessible to average users, and the
  programs and interfaces resulting from these
  efforts are the focus.
• A possible scenario of information retrieval
• Found a paper in PubMed which describes a gene in
  GenBank
• May want to know the protein coded from the gene
• May want to know its 3D structure recorded in
  structure database
• Entrez at NCBI (http//www.ncbi.nlm.nih.gov/Entrez
  )

15

• Visualization tools
• Cn3D

16
Informatics Techniques in Bioinformatics

• Databases
• Building
• Querying
• String comparison
• Text search
• Sequence alignment
• Significance statistics
• Pattern finding
• AI/Machine Learning
• Data mining

• Geometry
• Robotics
• Graphics
• 3D matching
• Physical simulation
• Numerical analysis
• Simulation
• Visualization

17

• Microarrays for Genomic Studies
• Why is it important?
• Its good to have information of (all) genes that
  are present in a genome.
• Genetics is the study of the interactions among
  individual genes in an organism.
• Biologists want to know the interplay of all
  genes simultaneously.
• This leads to the need for high-throughput and
  large-scale technologies.

18

• Gene Expression Studies
• Pattern of genes expressed in a cell is
  characteristic of its current state.
• Virtually all differences in cell state are
  correlated with changes in mRNA levels of many
  genes.
• Expression patterns of genes may provide clues to
  their functions by comparison.

19

• DNA Microarray Technology
• Initially conceived of to detect expression of
  1000s genes simultaneously.
• Potential applications
• Identification of complex genetic diseases
• Drug discovery
• Differing expression of genes over time between
  tissues or disease state
• Potential impacts
• Preventive medicine
• Ability to subtype diseases, and design drugs
  based on their causes instead of symptoms.
• Personalized drugs
• Rapid diagnosis

20

• Microarray Technology
• Basic technology is the same
• DNA sequence complementary to the gene of
  interest is generated and then laid out in
  microscopic quantities on solid surfaces at
  predefined positions
• DNAs from samples are washed out over the
  surface, only complementary DNAs are left due to
  binding
• Presence of bound DNAs is detected by
  fluorescence following laser excitation

21

• Microarray Fabrication
• Photolithography
• Photomasks direct DNA synthesis
• Add one base to growing chain at a time
• Ink-jetting
• Utilize ink-jet printing to dispense
  sub-nanolitre volume of reagent to defined
  positions
• Microspotting
• Robot with a printhead
• What DNA sequences are laid on surface
• A series of DNA fragments vs. complete gene
  sequence

22

• How it works
• Fluorescent samples are prepared from two mRNA
  sources to be compared
• Cy3 (green) used for one sample
• Cy5 (red) used for the other sample
• Samples are mixed and washed over the microarray
• The microarray is then scanned by a laser scanner
  to detect fluorescent level.
• The ratio of Cy3 to Cy5 (G/R) is calculated for
  each array element.
• Relative intensity of G/R is a reliable measure
  of the relative abundance of specific mRNA in
  each sample.

23

• Microarray Image

24

• ??????????
• ??????????????????
• ??????????????
• ????28????????,????????????,????????????,????????
  
• ????2.7????????????,???????,??B-Korea,????2025??
  ?,?????????????????
• ??????????(????)??????,?????????????????
• ?????1996???Bioinformatics Center,???????????680??
  ???????????????????
• ???????????????????O???,?????????????????????,????
  ??????????????????
• ??????????????????,?????????????????????,?????????
  ??????,????????????????,?????????,?IBM????????????
  ??

25

• ????????
• ?????????
• ????????,????????,???????
• ???,????????????????,???????
• ???????????(NIH)???,??????20??????????????
• ???????????
• ?????,??????????????????????,??????????,??????????
  ?,??????????????,??????????,???????

26

• ??????????
• ??????????????,???????
• ??????
• ???????????
• ????????,?????????????
• ????????,????????????????

27
Gene Regulation Analysis

• Motivation
• Microarray technology provides a global view of
  changes in gene expression on a genomic scale.
• Obtain temporal patterns of gene expression
• What else we may want to know about ?
• Consensus sequence motifs
• Correlation between motifs
• What cause genes to behave differently
• To learn beyond how genes behave over time
• Look into control regions of genes
• Find hypotheses

28
Gene Regulation Analysis

• Method
• Use Affymetrix GeneChip machine to collect gene
  expressions.
• Cluster genome based on temporal patterns, e.g.,
  slope, distance, etc.
• For each cluster, use motif-finding algorithm to
  find motifs.
• Given multiple clusters of interest, transform
  raw sequences into higher-level representation.
• Apply constructive induction to find motif
  interactions.
• Apply inductive learner to generate hypotheses.

29
(No Transcript)
30
RNA Secondary Structure Prediction Basics

• Like protein secondary structure, RNA secondary
  structure can be viewed as an intermediate step
  in the formation of a 3D structure.
• In predicting RNA secondary structure, several
  simplifying assumptions are usually made.
• The most likely structure is similar to the
  energetically most stable structure.
• The energy associated with any position in the
  structure is only influenced by local sequence
  and structure. most reliable when used for
  standard Watson-Crick base pairs and single G/U
  pairs surrounded by Watson-Crick pairs.
• The structure is assumed to be formed by folding
  of the chain back on itself in a manner that does
  not produce any knots.

31
Types of RNA Secondary Structure Prediction
Methods

• Based on objective functions
• Free energy minimization
• Covariance analysis from sequence comparison
• Based on number of RNA sequences for which to
  predict
• Single-sequence prediction
• To find the possible folding of a single RNA
  sequence
• Multiple-sequence prediction
• To find a global structure alignment for a set of
  RNA sequences
• To find common structure elements within a set of
  RNA sequences

32
Motif Prediction vs. Concept Learning

• Target concept common motifs
• Training examples biosequences
• Motif prediction as supervised learning
• Positive examples
• a given set of coregulated RNAs
• Negative examples
• the same number of sequences randomly generated
  based on the observed frequencies of sequence
  alphabet in positive examples.
• Target concept
• The common structural motifs that can be used to
  distinguish the given coregulated RNAs from the
  random sequences.

33
GPRM Genetic Programming for RNA Motifs

• Focus on finding Watson-Crick complementary
  basepairs
• C-G and A-U
• RNA secondary structures are typically formed by
  basepairing interactions.
• Three components of GPRM
• Population of putative structural motifs
• Fitness function of motifs
• Genetic operators that simulate the natural
  evolution process of motifs

34
Representing Individuals in A Population

• Each individual in a population is a putative
  motif
• Structural motif description
• Watson-Crick complementary segments
• Non-pairing segments

35
Fitness Function

• Interested in those motifs that can reflect the
  characteristics conserved in a family of
  coregulated RNAs
• Assign higher values to those motifs commonly
  shared by the given family of RNAs, and rarely
  contained in random RNA sequences.
• We define the fitness function as

36
Genetic Operators

• Reproduction
• Pass the better half of the population to the
  next generation
• Accelerate the reproduction process
• Mutation
• If a complementary segment is picked, its segment
  length and corresponding pairing segment are both
  randomly changed.
• If a non-pairing segment is selected, then only
  its length is randomly modified.
• Crossover
• Exchange segment configuration between two
  putative motifs.
• Either a pair of complementary segments or a
  non-pairing segment is randomly chosen for
  exchange.

37
(No Transcript)
38
(No Transcript)
39
Reconstruction of Transcriptional Regulatory
Networks

• Various Genome Projects produce a sufficient
  amount of sequence data
• Microarray technologies generate a large amount
  of gene expression data
• Large-scale and high-throughput sequence data and
  expression profiles are considered one of the
  most promising techniques to reconstruct gene
  networks
• Computational analysis and reconstruction of
  genetic regulatory networks is now feasible
• Goal Combining different types of information to
  model transcriptional regulatory networks for
  genes and TFs of interest

40
Methods

• A Bottom-up Approach
• transcription modules ? transcription network
• Transcription module a functional unit
  consisting of a TF, target genes, genes producing
  this TF

41
Objective

• Incorporate several hypotheses and analyze
  multiple data sources to enhance the performance
  in predicting transcription modules and construct
  transcription networks from these modules.
• The transcription binding site information
• Expression profile similarity of potential
  co-regulated target genes
• Correlation between expression profile of the
  genes producing the TF and that of those
  regulated

42
Background Hypotheses

• The development of large-scale expression
  monitoring and the availability of complete
  genome sequence allow the refinement of
  computational analysis.
• A candidate gene is considered a target gene of a
  particular TF if
• The upstream region of the gene contains the
  binding sites of the TF.
• The PEA associated with the gene is significantly
  small.
• Expression profile similarity of potential
  co-regulated target genes
• The PF associated with the gene is significantly
  small.
• Correlation between a target gene and the genes
  producing TF

43
Synergy of Binding Sites, PEA, PF

• Binding sites
• IUPAC/IUB code
• Matched against upstream to provide preliminary
  candidate cis-regulatory sequences
• Problem false positives
• Combination of sequence similarity and expression
  phenotype to reduce false positives
• PEA (Probabilistic Element Assessment)
• PF (P-value of F test)

44
Synergy of PF and PEA

• Varying PF and PEA (0.4 10-6)
• PF-PEA combination against PF or PEA alone
• Over all 27 TFs, appropriate PF-PEA threshold
  combo outperforms the best single PF or PEA
• Prove the synergy of PF-PEA combination

45
Results Reconstruction of Transcription Networks

• Goal to reconstruct the transcriptional
  regulatory network for a given set of TFs and
  genes of interest
• TFs of interest MCM1, ACE2, SWI5, SBF, MBF
• Parameter Setting PF threshold 0.03 and PEA
  threshold 0.65
• Case 1 CLB1, CLB2, SWI5, ACE2, CDC5, CLN3, SWI4,
  FAR1, RME1, SIC1, CDC6, CLN1, CLN2, CLB5 and CLB6
  related to Cell Cycle
• Case 2 randomly pick 130 mitotic cell
  cycle-related genes from CYGD

46
Advantages of Combinatorial Approach

• Different metrics cover different background
  knowledge
• Exploit more information to avoid false positives
  when building transcription modules
• These metrics complement each other by
  characterizing different biological activities
• e.g. Similar expression profiles among
  co-regulated genes and the association between
  regulators and the target genes
• More robust
• e.g., In case TF binding sites are unavailable,
  regression analysis can still be applied to
  identify reasonable transcription modules.

47
(No Transcript)