1. Microarray
Yuki Juan
NTUST
May 26, 2003

2. Content Biology background of microarray Design of microarray
The workflow of microarray
Image analysis of microarray
Data analysis of microarray
Discussion

3. The Biology Background of Microarray
The central dogma of life forms
DNA
RNA
Monitoring the expression of genes

4. Central Dogma DNA Replication RNA Transcription Protein Translation
--ACGCGA--
--TGCGCT--
RNA Transcription
--UGCGCU--
Protein Translation
--CYSALA--

5. DNA
replication
transcription
translation
DNA
RNA
Protein

6. DNA The double helix Nucleotide Base pair Oligonucleotide stable
A, T, G, C
Base pair
A – T
G – C
Oligonucleotide
short DNA (tens of nucleotides, or bps) (

7. DNA Strand DNA has canonical orientation read from 5’ to 3’
antiparallel: one strand has direction opposite to its complement’s
5’ … TACTGAA … 3’
3’ … ATGACTT … 5’

8. Hydrogen Bond Makes DNA Binding Specifically5’ 3’ 5’ 3’

9. Hydrogen Bond Makes DNA Binding Specifically
The force between base pair is hydrogen bond, This force let
A-T(U), C-G can specifically match together.

10. RNA
replication
transcription
translation
DNA
RNA
Protein

11. RNA Types messenger RNA ribosomal RNA (rRNA) transfer RNA (tRNA)
Gene is expressed by transcribing DNA
into single-stranded mRNA

12. RNA (Detailed) (

13. Reverse Transcription
replication
transcription
translation
DNA
RNA
Protein
Reverse Transcription
By reverse transcriptase, we can convert RNA into cDNA.

14. The Southern Blot
Basic DNA detection technique that has been used for over 30 years, known as Southern blots:
A “known” strand of DNA is deposited on a solid support (i.e. nitocellulose paper)
An “unknown” mixed bag of DNA is labelled (radioactive or flourescent)
“Unknown” DNA solution allowed to mix with known DNA (attached to nitro paper), then excess solution washed off
If a copy of “known” DNA occurs in “unknown” sample, it will stick (hybridize), and labeled DNA will be detected on photographic film

15. mRNA Represent Gene Function
When measure the level of a mRNA, we are monitoring the activity of a gene.
Thus, if we can understand all the level of mRNAs, we can study the expression of whole genome.
Microarray takes the advantage of getting over of blotting data in a single experiment, which makes monitoring the genome activity possible.

16. Content Biology background of microarray Design of microarray
The workflow of microarray
Image analysis of microarray
Data analysis of microarray
Discussion

17. Design of Microarray Microarray in different context
The idea of microarray
Main type of array chips

18. mRNA Levels Compared in Many Different Contexts
Different tissues, same organism (brain v. liver)
Same tissue, same organism (tumor v. non-tumor)
Same tissue, different organisms (wt v. mutant)
Time course experiments (development)
Other special designs (e.g. to detect spatial patterns).

19. Idea of Microarray Cell A Cell B Labeled cDNA from geneX
Hybridizaton to chip
Spot of geneX with complementary sequence of colored cDNA
This spot shows red color after scanning.

20. Over 10,000 Hybridization Could Be Down at One Time

21. Several Types of Arrays
Spotted DNA arrays
Developed by Pat Brown’s lab at Stanford
PCR products of full-length genes (>100nt)
Affymetrix gene chips
Photolithography technology from computer industry allows building many 25-mers
Ink-jet microarrays from Agilent
25-60-mers “printed directly on glass slides
Flexible, rapid, but expensive

22. Array Fabrication Spotting
Use PCR to amplify DNA
Robotic "pen" deposits DNA at defined coordinates
approximately 1-10 ng per spot
Experimentation with oligos (40, 70 bp)

23. This machine can make 48 microarrays simultaneously.

24. Array Fabrication Photolithography
Light activated synthesis
synthesize oligonucleotides on glass slides
107copies per oligo in 24 x 24 um square
Use 20 pairs of different 25-mers per gene
Perfect match and mismatch

25. Array Fabrication Photolithography

26. Affymetrix Microarrays
Raw image
1.28cm
50um
~107 oligonucleotides,
half perfectly match mRNA (PM),
half have one mismatch (MM)
Raw gene expression is intensity difference: PM - MM

27. Agilent cDNA microarray and oligonucelotides microarray
Agilent delivering printed 60-mer microarrays in addition to 25-mer formats.
The inkjet process uses standard phosphoramidite chemistry to deliver extremely small volumes (picoliters) of the chemicals to be spotted.

28. Content Biology background of microarray Design of microarray
The workflow of microarray
Image analysis of microarray
Data analysis of microarray

29. The Workflow of Microarray
sample
Plate
Plate Preparation
RNA extraction
Array Fabrication
cDNA synthesis
and labeled
Array
Hybridization
Labeled cDNA
Hybridized Array
Scanning

30. cDNA Synthesis And Directly Labeling

31. Cy3 and Cy5 cDNA Hybridization On To The Chip
e.g. treatment / control
normal / tumor tissue
Sample loading
1.Loading from the corner of the cover slip
It is time consuming and easily producing bubbles. 1
2. Loading sample at the center of array then put the slip smoothly
Faster, and have lower chance of bubble producing then the last one. 2
Sample loading
3. Loading sample at the side of the array then put the slip on.
Solution would attach to the slip right after the slip contact with it, and would diffuse with the movement of slip when we slowly move down. 3
Sample loading

32. Scan
Green: down regulate
Red: up regulate
Yellow: equal level

33. Content Biology background of microarray Design of microarray
The workflow of microarray
Image analysis of microarray
Data analysis of microarray
Discussion

34. Image analysis To find a spot Convert feature into numeric data
Image normalization

35. The Algorithms
1. Find spots: Finds the location of each spot on the microarray.
2. Cookie cutter algorithm:
(1).Suppose the distribution of pixels vs intensity is Gaussian curve
(2).Using SD or IQR to identify the feature and background of each spot
(3).Calculates statistics for the pixel population

36. Interquartile Range(IQR)D
K=IQR/2
1.42 IQR
Boundary for rejection
25%
50%
75%
Boundary for rejection
IQR

37. Feature or cookie D
Local background
Exclusion zone

38. Data Quality Irregular size or shape Irregular placement Low intensity
Saturation
Spot variance
Background variance
miss alignment
artifact
indistinguishable
saturated
bad print

39. Convert Feature Into Numeric Value
Green background
Green b.g.-corrected
Red b.g.-corrected
(R. b.g.-c)/(G. b.g.-c)
Red intensity
Green intensity
Systematic name
Red b.g.
Gene function

40. Data Normalization Normalize data to correct for variances
Dye bias
Location bias
Intensity bias
Pin bias
Slide bias
Control vs. non-control spots

41. Data Normalization Uncalibrated, red light under detected
Calibrated, red and green equally detected

42. Data Normalization Assumptions Overall mean average ratio should be 1
Most genes are not differentially expressed
Total intensity of dyes are equivalent

43. Intensity Dependent Normalization

44. After Normalization

45. Additional Normalization
Pin dependent
Similar to intensity dependent fit.
Compute individual lowess fits for each pin group
Within slide normalization
After pin dependent normalization, log ratios for each pin are centered around 0
Scale variance for each pin
Uses MAD (median absolute deviation)

46. Additional Normalization
Dye swap
Combine relative expression levels without explicit normalization
Compute lowess fit for
log2(RR’/GG’)/2 vs. log2(A + A’)/2
Normalized ratio is
log2(R/G) - c(A)
where c(A) is the lowess prediction

47. Content Biology background of microarray Design of microarray
The workflow of microarray
Image analysis of microarray
Data analysis of microarray
Discussion

48. Data analysis Data filtering Fold change analysis Classification
Clustering
Future direction

49. Microarray Data Classification
Microarray chips
Images scanned by laser
Gene Value
D26528_at
D26561_cds1_at
D26561_cds2_at
D26561_cds3_at
D26579_at
D26598_at
D26599_at
D26600_at
D28114_at
Datasets
New
sample
Data Mining
and analysis
Prediction:

50. The Threshold of Spots
Filtering - remove genes with insufficient variation
Remove insufficient spot:
saturated, None uniform, too high background…
Remove extreme signal:
e.g. MaxVal - MinVal < 500 and MaxVal/MinVal < 5
Statistical filtering (e.g. p-value<0.01)
biological reasons
feature reduction for algorithmic

51. Microarray Data Analysis Types
Different gene expression
Fold change analysis
Classification (Supervised)
identify disease
predict outcome / select best treatment
Clustering (Unsupervised)
find new biological classes / refine existing ones
exploration …

52. Differential Gene Expression
n-fold change
n typically >= 2
May hold no biological relevance
Often too restrictive
2 expression
Calculate standard deviation 
Genes with expression more than 2 away are differentially expressed

53. Fold Changes-Scatter Plot21

54. Fold Changes Table 23

55. Classification: Multi-Class
Similar Approach:
select top genes most correlated to each class
select best subset using cross-validation
build a single model separating all classes
Advanced:
build separate model for each class vs. rest
choose model making the strongest prediction

56. Popular Classification Methods
Decision Trees/Rules
find smallest gene sets, but also false positives
Neural Nets -
work well if number of genes is reduced
SVM
good accuracy, does its own gene selection, hard to understand
K-nearest neighbor - robust for small number genes
Bayesian nets - simple, robust

57. Multi-class Data Example
Brain data, Pomeroy et al 2002, Nature (415), Jan 2002
42 examples, about 7,000 genes, 5 classes
Selected top 100 genes most correlated to each class
Selected best subset by testing 1,2, …, 20 genes subsets, leave-one-out x-validation for each

58. Classification – Other Applications
Combining clinical and genetic data
Outcome / Treatment prediction
Age, Sex, stage of disease, are useful
e.g. if Data from Male, not Ovarian cancer

59. Clustering Goals Find natural classes in the data
Identify new classes / gene correlations
Refine existing taxonomies
Support biological analysis / discovery
Different Methods
Hierarchical clustering, SOM's, etc

60. SOM clustering SOM - self organizing maps Preprocessing
filter away genes with insufficient biological variation
normalize gene expression (across samples) to mean 0, st. dev 1, for each gene separately.
Run SOM for many iterations
Plot the results

61. SOM & K Mean By GeneSpring27

62. Hierarchical Clustering
The most popular hierarchical clustering method used in microarray data analysis is the so called agglomerative method
works with the data in a bottom-up manner.
Initially, each data point forms a cluster and the algorithm works through the cluster sets by repeatedly merging the two which are the most similar or have the shortest distance.
algorithm involves the computation of the distance or similarity matrix
O(N^2) complexity and thus is not very efficient.

63. Hierarchical clustering

64. Future directions
Algorithms optimized for small samples (the no. of samples will remain small for many tasks)
Integration with other data
biological networks
medical text
protein data
cost-sensitive classification algorithms
error cost depends on outcome (don’t want to miss treatable cancer), treatment side effects, etc.

65. Integrate biological knowledge when analyzing microarray data (from Cheng Li, Harvard SPH)
Right picture: Gene Ontology: tool for the unification of biology, Nature Genetics, 25, p25

66. Content Biology background of microarray Design of microarray
The workflow of microarray
Image analysis of microarray
Data analysis of microarray
Discussion

67. Microarray Potential Applications
Biological discovery
new and better molecular diagnostics
new molecular targets for therapy
finding and refining biological pathways
Mutation and polymorphism detection
Recent examples
molecular diagnosis of leukemia, breast cancer, ...
appropriate treatment for genetic signature
potential new drug targets

68. Microarray Limitations
Cross-hybridization of sequences with high identity
Chip to chip variation
True measure of abundance?
Does mRNA levels reflect protein levels?
Generally, do not “prove” new biology - simply suggest genes involved in a process, a hypothesis that will require traditional experimental verification.
What fold change has biological relevance?
Need cloned EST or some sequence knowledge -- rare messages may be undetected
Expensive!! Not every lab can afford experiment repeat.
The real limitation is Bioinformatics

69. Additional Information
Review papers on microarray
Genomics, gene expression and DNA arrays (Nature, June 2000)
Microarray - technology review (Natural Cell Biology, Aug. 2001)
Magic of Microarray (Scientific American, Feb. 2002)
Molecular biology tutorial

70. Biological data retrieval systems: Entrez http://www. ncbi. nlm. nih
A retrieval system for searching a number of inter-connected databases at the NCBI. It provides access to:
PubMed: The biomedical literature (Medline)
Genbank: Nucleotide sequence database
Protein sequence database
Structure: three-dimensional macromolecular structures
Genome: complete genome assemblies
PopSet: population study data sets
OMIM: Online Mendelian Inheritance in Man
Taxonomy: organisms in GenBank
Books: online books
ProbeSet: gene expression and microarray datasets
3D Domains: domains from Entrez Structure
UniSTS: markers and mapping data
SNP: single nucleotide polymorphisms
CDD: conserved domains
2. Entrez allows users to perform various searches.