Basic Aspects of Microarray Technology and Data Analysis (UEB-UAT Bioinformatics Course - Session 3.2 - VHIR, Barcelona)

Bioinformàtica per a la Recerca Biomèdica Ricardo Gonzalo Sanz

ricardo.gonzalo@vhir.org 20/05/14

Hospital Universitari Vall d’Hebron Institut de Recerca - VHIR

Institut d’Investigació Sanitària de l’Instituto de Salud Carlos III (ISCIII)

Basic aspects of Microarray technology

Affymetrix microarrays manufacture.

2

3

4

5

6

Microarray experiment workflow.

Quality Controls.

Different types of Affymetrix arrays.

1 Introduction

Different types of arrays. Manufactoring. DNA/RNA/Protein

1 Introduction

reproducibility

only show you what you’re looking for

what about ‘indels’, inversions, translocations...

accuracy

sensitivity

1 Introduction

1 Introduction

RNA-Seq was superior in detecting low abundance transcripts

also better detecting differentiating biologically isoforms

RNA-Seq demonstrated a broader dynamic range than microarray.

1 Introduction

• In molecular biology exist a lot of techniques to measure the gene expression

(Northern blot)

• Main characteristic from the microarrays discovery (Schena et al. (1995)

Science 270:467-70), was not what could be measured, instead the quantity of

simultaneous measures that could be done.

• Pre microarrays time: study of genes was one by one

• Post microarrays time: all the genes together.

1 Introduction

• But.... what is a microarray in few words?

DNA fixed to a solid surface (nylon, silica, glass,...)

RNA “problem” is labeled and have to bind to DNA

fixed in the solid surface in an specific way.

DNA binded usually is called “probe”

Labeled RNA usually is called “target”

Important to know in advanced...

1 Introduction

• Microarrays are usually hypothesis-generating:

They highlight specific genes or features that are particularly

interesting for follow-up experiments.

An exception would be the biomarkers discovery studies.

• This does not reduce the importance of experimental design

2

Two color microarrays (cDNA)

• Usually probes are long (20nt)

• Probe is fixed to a glass

• Labeling is with two fluorocrom (Cy3/Cy5).

• Direct comparison of the two samples due

to they are hybridized in the same array.

• Each gene appear few times in the array

• Long probes facilitate crosshybridization

• Not very good reproducibility.

Different types of arrays. Manufactoring. DNA/RNA

2

One color microarrays

• Short probes (20-25 nt)

• Target is labeled with only one fluorocrom

• Only one sample is hybridized in each array.

• Each gene is represented by a lot of probes

in the array

Different types of arrays. Manufactoring. DNA/RNA

2 Different types of arrays. Manufactoring. DNA/RNA

• DNA Polymorphism (GWAS)

• Transcription Factors

• Resequencing

• Cytogenetics

• Expression

• Alternative splicing

• microRNA

DNA RNA

2 Different types of Affymetrix arrays.

3’ 5’

3’ IVT Arrays

• Biased measurement of the gene expression

• Array more used in the literature. A lot of species present.

Only genes with polyA tail and good 3’ site will

be amplified and will have the chance of

hybridize correctly.

2 Different types of Affymetrix arrays.

3’ 5’

Gene Arrays

Exon Arrays

Gene/Exon Arrays

• Gene arrays are the most used (good quality and price ratio)

• Gene arrays 2.0 more updated library and also includes lncRNAs

2 Different types of expression arrays.

•153 organisms in the array (human, mouse, rat, canine, ….)

•100% miRBase v17

•2.216 snoRNAs and scaRNAs (human small nuclear RNAs)

•Low inputs amounts (130 ng total RNA)

•2.999 probe sets unique to pre-miRNA hairpins

•Able to differentiate pre and mature miRNAs

•Useful for FFPE samples

miRNA

2 Different types of expression arrays.

HTA array

Affymetrix microarrays manufacture. 3

Photolitografy

Affymetrix microarrays manufacture. 3

5 Microarray experiment workflow

5 Microarray experiment workflow

5 Microarray experiment workflow

6 Quality Controls

6 Quality Controls

6 Quality Controls

Length of amplified cRNA

6 Quality Controls

Length of fragmented cRNA

Bioinformàtica per a la Recerca Biomèdica Ricardo Gonzalo Sanz

ricardo.gonzalo@vhir.org 20/05/14

Hospital Universitari Vall d’Hebron Institut de Recerca - VHIR

Institut d’Investigació Sanitària de l’Instituto de Salud Carlos III (ISCIII)

Basic aspects of Microarray Data Analysis

Filtering

2

3

4

5

6

Statistical inference of diferential expression

Clustering

Normalization

1 Introduction. Experimental design

Quality control

7

8

Annotation

Biological interpretation

1 Introduction. Experimental design

1 Introduction. Experimental design

1 Introduction. Experimental design

1 Introduction. Experimental design

1 Introduction. Experimental design

1 Introduction. Experimental design

Microarrays Analysis

Workflow

2 Quality Control

2 Quality Control

Was the experiment a success???

• Microarray experiments generate huge quantitites of data

• Standard statistical approach use plots to check the quality

show all data together

highlight structures

may help to detect problems (“unusual patterns”)

It is hard to decide if things “seem to be

all right” just by looking at the numbers.

2 Quality Control

Diagnostics plots for microarrays:

• Microarray data usually considered at two levels

1. Low level. Data directly coming from the scanner

2. High level. Processed from low level data. Expression values,

normalized or not.

• Some plots are specific for some type of arrays or for some level

2 Quality Control

Diagnostics plots for microarrays:

1. Low level:

Layout image

Degradation plots (only in 3’IVT)

Histogram/density plots

PCA, Boxplot

2. High level:

MA plots

Model based plots (NUSE,RLE,)

PCA, Boxplot

2 Quality Control

Diganostics plots for microarrays. Low level. Layout image.

2 Quality Control

Diagnostic plots for microarrays. Low level. RNA degradation plot (3’IVT arrays)

2 Quality Control

Diagnostics plots for microarrays. Low level. Histogram/density Plot

2 Quality Control

Diagnostics plots for microarrays. Low level. Boxplot

2 Quality Control

2 Quality Control

Diagnostics plots for microarrays. Low level. PCA

2 Quality Control

Diagnostics plots for microarrays. Low level. PCA

2 Quality Control

2 Quality Control

Diagnostics plots for microarrays. High level. RLE

2 Quality Control

2 Quality Control

Diagnostics plots for microarrays. High level. NUSE

2 Quality Control

Diagnostics plots for microarrays. High level. MA plots

• MA plots allow pair wise comparison of log-intensity of each array to a

reference array and identification of intensity-dependent biases.

• The Y axis of the plot contains the log-ratio intentsity of one array to the

reference median array, which is called “M” while the X axis contains the

average log-intensity of both arrays – called “A”.

• The probe levels are not likely to differ a lot so we expect a MA plot centered

on the Y=0 axis from low to high intensities.

2 Quality Control

Diagnostics plots for microarrays. High level. MA plots

2 Quality Control

3 Normalization

The goal of normalization is to adjust for the effects that are due to variations in the

technology rather than the biology.

3 Normalization

3 Normalization

3 Normalization

4 Filtering

• In a microarray experiment only a few hundreds/thousand of genes change their

expression due to the different conditions

•Researcher is interested in keeping the number of tests/genes as low as possible

while keeping the interesting genes in the selected subset.

•If the truly diferentially expressed genes are over-represented among those

selectec in the filtering step, the FDR associated with a certain threshold of the

statistic test will be lowered due to the filtering.

Genes that do not change introduce

noise, therefore is better not to be

present when the statistical analysis is

done

4 Filtering

Exists different types of filtering:

• Annotation features (specific):

Specific gene features (i.e. GO term, presence of transcriptional regulative

elements in promoters, etc.)

Data derived from IPA

• Signal features (non specific)

% intensities greater of a user defined value

Interquantile range (IQR) greater of a defined value

4 Filtering

Signal filtering: This technique has as its premise the removal of genes that are

deemed to be not expressed or unchanged according to some specific criterion that

is under the control of the user.

5 Statistical inference of diferential expression

• Indirect comparisons: 2 groups, unpaired

• Direct comparsions: 2 groups. paired

5 Statistical inference of diferential expression

Limma package (Gordon Smith)

5 Statistical inference of diferential expression

5 Statistical inference of diferential expression

5 Statistical inference of diferential expression

5 Statistical inference of diferential expression

6 Clustering

Types:

Supervised clustering try to find the best partition for data that belong to a

know set o classes

Unsupervised clustering try to define the number and the size of the classes

in which the transcription profiles can be fitted in.

6 Clustering

6 Clustering

Hierarchical Clustering (HCL)

• HCL is an agglomerative /divise clustering method.

• The iterative process continues until all groups are

connected in a hierarchical tree.

• Samples more similar between them are closed.

6 Clustering

7 Annotation

8 Biological interpretation

Gene Ontology

8 Biological interpretation