Transcript - Lorentz Center

Microarray Analysis Software
Maximiliano Corredor
Institute of Biology, Leiden University
Steps of a
Microarray
Experiment
Genomic sequence /
EST library sequence
RNA
RT
Annotation
cDNA
labeling
cDNA-Cy3 / -Cy5
hybridization
Image
Processing
Statistical
Analysis
Probe design
Bioinformatic steps of MA experiments
• Probre design
• Image processing (with QC)
• Normalisation (with QC)
• Statistical analysis and data mining
• Database management
Probe design software
•
Array Designer - a software that can design hundreds of primer for DNA or
oligonucleotide microarrays, product of Premier Biosoft.
http://www.premierbiosoft.com/dnamicroarray/index.html
•
OligoArray2 - a free software that computes gene specific oligonucleotides for
genome-scale oligonucleotide microarray construction.
http://berry.engin.umich.edu/oligoarray2/
•
OligoWiz2 Server - server for designing oligonucleotide probes for microarrays.
http://www.cbs.dtu.dk/services/OligoWiz2/
•
ProbeWiz Server - The CBS ProbeWiz WWW server predicts optimal PCR
primer pairs for generation of probes for cDNA arrays.
http://www.cbs.dtu.dk/services/DNAarray/probewiz.php
•
Primer3 - a common used software for designing primers for microarray
construction.
http://frodo.wi.mit.edu/primer3/primer3_code.html
Image processing
• Addressing: estimate location of spot
centers
• Segmentation: classify pixels as
foreground or background
• Information Extraction: for each spot
on the array and each channel
• Foreground intensities
• Background intensities
• quality measures
Image processing software
GenePix Pro (Axon Instruments) for Windows
• Spot identification, scatter plot, histogram, normalization, quality control
http://www.moleculardevices.com/pages/software/gn_genepix_pro.html
ScanArray (PerkinElmer) for Windows
• Quantitation, spot quality measures and normalization
http://las.perkinelmer.com/Catalog/default.htm?CategoryID=Analysis+Software
ScanAlyze (Eisen's lab, Lawrence Berkeley National Lab (LBNL). For Windows
• Process fluorescent images of microarrays. Semi-automatic definition of grids
and complex pixel and spot analyses. Free for academic
http://rana.lbl.gov/EisenSoftware.htm
TIGR Spotfinder (TIGR) for Windows
• Spot identification; Microarray image processing. Free
Image processing with GenePix
QC: Background substraction
• Background arises from glass autofluorescence,
dust particles or washing defects
• BG and specific hybridisation are assumed
additive (but look at the image!!)
• Low background can be substracted from the
average intensity of the spot.
• High background features should be removed
from analysis: artificial saturation may occur and
therefore the maximum measure is not the
addition of background and real specific
intensity.
• Features with high negative intensities after
background substraction (like those of the
image) should also be removed.
• Features with background similar to spot
intensity will give a normal distribution centered
in 0 intensity and can therefore be considered
absent.
Background correction
• Different types of background substraction
• Possibility of flagging features that don’t match our QC criteria:
- high background intensity
- % of pixels above
background
- background higher
than foreground
QC: Histogram and scatterplot
The intensities should follow a normal distribution with:
– Natural lower limit: only positive intensities exist (minimum RNA
concentration is 0)
– Long tail to the higher intensities
– Artificial upper limit: saturation of detector and/or TIFF file. This can cause
an accumulation of points at the highest intensity
This effect can also be observed in the scatterplot
QC: Std. Dev. vs. Avg
Good spots should be homogenous: low
standard deviation
• Linear correlation std. dev. vs average
• Higher std dev = variability within spot
• Lower std dev = uniformity within spot
(saturation)
Sources of technical variability
• Chip production
• efficiencies of
-RNA extraction
-reverse transcription
-labeling
-photodetection
SYSTEMATIC
• Calibration can correct
for them
•
•
•
•
•
PCR yield
DNA quality
spotting efficiency,
spot size
cross-/unspecific
hybridization
• stray signal
STOCHASTIC
• Error model normalization
Normalisation
• Several assumptions:
– Normal distribution of intensities
– All channels behave equally
• Centering and scaling:
– Intensities are transformed in a way that the averages and ranges are
the same (and therefore comparable)
• Within hyb normalisation:
– In two channel data, both channels are centered and scaled.
– More complex normalisations may be needed in order to ensure
linearity along all intensities range.
• Between hybs normalisation:
– Every time that two or more different chips are going to be compared,
it’s necessary that all of them are centered and scaled
– Normalisation should be made taking into account the experimental
design; error model must include distinction between experimental units,
biological replicates and technical replicates
Normalisation software
• Basic normalisation within hybridisation is possible in
GenePix
• Acuity includes more advanced normalization algorithms
(Lowess, etc)
• Rosetta implements several pipelines for normalization
– Within hybs when uploaded to the database, using manufacturer
indications for developing their error-models (providing therefore
with p-values)
– Between hybs when compared to each other (centering and
scaling)
QC: M vs A
• M stands for Log(Ratio); A is the product of the
Log(Intensity) of both channels.
• If the two channels behave symmetrically, everything is
OK. Otherwise, we may have dye bias
• It is very common to find such deviations in the tails of
the distribution (lowess normalisation can help here).
QC: M vs A
•
•
Before normalisation (left), average ratio was higher than 0.
Intensity saturation of one channel produces skewed tail. This effect is not
removed with normalisation, requires calibration of the image acquisition (or
elimination of saturated spots from analysis)
QC and basic statistics software
• Some image processing packages include basic statistics functions,
like GenePix
• Numerous stand-alone programs and plug-ins or scripts for more
general statistical packages, like R/Bioconductor, Matlab, SPSS, MS
Excel…
http://ihome.cuhk.edu.hk/~b400559/arraysoft_statistics.html
• All microarray analysis packages include this functions and many
more
Database systems
Acuity (Axon Instruments)
• Runs on Windows 2000/XP client; Windows 2000 server (recommended)
• Stores data in relational database, Microsoft SQL or Oracle
• Various visualization tools; normalization; hierarchical, k-means, k-medians
clustering with many different similarity metrics, SOM, PCA, gene shaving.
• Scripting engine for customizable analysis
http://www.moleculardevices.com/pages/software/gn_acuity.html
ArrayDB (NHGRI)
• Html/ linux or Unix
• Analyzed expression data stored in a relational database
• a software suite that provides an interactive user interface for the mining
and analysis of microarray gene expression data.
http://genome.nhgri.nih.gov/arraydb/
Database systems
BASE (BioArray Software Environment) Department of Oncology, Lund
University
• Linux server, MySQL, web client
• Manages biomaterial information, raw data and images, and provides
integrated and "plug-in"-able normalization, data viewing and analysis tools.
• The system also has array production LIMS features; support MIAME and
MAGE-ML
Rosetta Resolver (Rosetta Biosoftware)
• JAVA/ UNIX with Oracle relational database
• The Rosetta Resolver system combines advanced analysis software, a
high-capacity database, and high-performance server framework in one
enterprise-wide tool.
Database systems
Stanford Microarray Database (SMD) package (Stanford University)
• Oracle server; web server; UNIX with Perl support
• SMD stores raw and normalized data from microarray experiments, as well
as their corresponding image files. In addition, SMD provides interfaces for
data retrieval, analysis and visualization.
http://genome-www5.stanford.edu//download/
Longhorn Array Database (Institute for Cellular and Molecular Biology,
University of Texas at Austin)
• Linux and PostgreSQL
• The Longhorn Array Database (LAD) is a MIAME compliant microarray
database. It is a fully open source version of the Stanford Microarray
Database (SMD)
http://www.longhornarraydatabase.org/
Rosetta Resolver
•
Excellent database
But requires dedicated staff to maintain
•
Ideal for institutions and big companies
Who are the only ones able to afford it
•
Includes a good set of statistical tools
But it isn’t very transparent
•
GUI user-friendly(ish)
•
Flexible advanced statistics available as visual
scripts and R implementation
However this requires deep knowledge of the
DB structure and some programming skills
•
Compatible with multitude of data formats
But hard to get info out of the system (no
MIAME yet)
Statistical Analysis and Data Mining
• Basic output of a microarray experiment is a list of genes
differentially transcribed. This can be obtained easily (Excel) from
the image processing.
• However the list is arbitrary: fold-change values are arbitrarily
chosen and there is no measure of the significance of the observed
difference: to do science we need statistics
• Many packages like Acuity, BASE and Rosetta Resolver combine
database and statistical analysis tools, but there are also many other
programs exclusively devoted to the statistical analysis of microarray
experiments:
http://ihome.cuhk.edu.hk/~b400559/arraysoft_mining_comprehensiv
e.html
Statistical analysis and
Data mining software
•
GeneSpring (Silicon Genetics) Analyze various array types, scatter plot,
cluster analysis, PCA, SOM, statistic tools, 2D, 3D plotting
•
J-Express (MolMine) Hierarchical clustering, K-means particional clustering,
Principal component anlaysis, Self-organizing maps, Profile similarity
search, Normalization and filtering, Raw data import, Project organization.
Free for academics
•
BioConductor, an open source software project providing infrastructure in
terms of design and software for analysing genomic data, some form of
graphical user interface for selected libraries. For other microarray related R
packages: http://ihome.cuhk.edu.hk/~b400559/arraysoft_rpackages.html
•
SpotFire (Spotfire) Hierarchical, bi-directional hierarchical and K-means
cluster analysis, PCA, profile search, coincidence testing, normalization, a
number of interactive plots for visualization of data, access GATC
databases
Basic plots and tables
Basic plots and tables
Classification tasks for microarrays
• Classification of SAMPLES
Generate gene expression profiles that can
(i) discriminate between different known cell types or
conditions, e.g. between tumor and normal tissue,
(ii) identify different and previously unknown cell types or
conditions, e.g. new subclasses of an existing class of
tumors.
• Classification of GENES
(i) Assign an unknown cDNA sequence to one of a set of
knowngene classes.
(ii) Partition a set of genes into new (unknown) functional
classes on the basis of their expression patterns across a
number of samples.
• Discriminant analysis: CLASSES KNOWN
• Cluster analysis: CLASSES NOT KNOWN
Cluster analysis
• Grouping a collection of objects into subsets or “clusters”, such that
those within each cluster are more closely related to one another than
objects assigned to different clusters.
• Two ingredients are needed to group objects:
– Distance measurement
– Clustering algorithm
• Clustering columns: grouping similar samples
• Clustering rows: grouping similarly expressed genes
Clustering of genes
•
Genes with similar patterns of expression
(synexpression groups) cluster together.
•
Synexpression groups may be functional
groups (this is a hypothesis that always
has to be tested).
Iyer et al., Science 1999
Clustering of samples
• Provided enough number of samples,
functional relationships might be
found
Golub et al. http://www.genome.wi.mit.edu/MPR
Discriminant analysis
Useful links
•
http://ihome.cuhk.edu.hk/~b400559/arraysoft.html
Comprehensive recopilation of information on microarray software
•
https://www.cs.tcd.ie/Nadia.Bolshakova/softwaretotal.html
Catalogue of microarray analysis software
•
http://genome-www5.stanford.edu/resources/restech.shtml
Stanford Microarray Database Software and Tools
•
http://www.tigr.org/software/microarray.shtml
The Institute for Genomic Research Microarray Software