Transcript slides
Probe selection for Microarrays
Considerations and pitfalls
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Probe selection wish list
Probe selection strategy should ensure
Biologically meaningful results (The truth...)
Coverage, Sensitivity (... The whole truth...)
Specificity (... And nothing but the truth)
Annotation
Reproducibility
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Technology
Probe immobilization
Oligonucleotide coupling
Synthesis with linker, covalent coupling to surface
Oligonucleotide photolithography
ds-cDNA coupling
cDNA generated by PCR, nonspecific binding to surface
ss-cDNA coupling
PCR with one modified primer, covalent coupling, 2nd strand
removal
Spotting
With contact (pin-based systems)
Without contact (ink jet technology)
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Technology-specific requirements
General
Not too short (sensitivity, selectivity)
Not too long (viscosity, surface properties)
Not too heterogeneous (robustness)
Degree of importance depends on method
Single strand methods (Oligos, ss-cDNA)
Orientation must be known
ss-cDNA methods are not perfect
ds-cDNA methods don’t care
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Probe selection approaches
Accuracy
Throughput
Selected Gene
Regions
Selected
Genes
ESTs
Cluster
Representatives
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Anonymous
Non-Selective Approaches
Anonymous (blind) spotting
Using clones from a library without prior sequencing
Only clones with interesting expression pattern are sequenced
Normalization of library highly recommended
Typical uses:
HT-arrays of ‘exotic’ organisms or tissues
Large-scale verification of Differential Display clones
EST spotting
Using clones from a library after sequencing
Little justification since sequence availability allow selection
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Spotting of cluster representatives
Sequence Clustering
For human/mouse/rat EST clones: public cluster libraries
Unigene (NCBI)
THC (TIGR)
For custom sequence: clustering tools
STACK_PACK (SANBI)
JESAM (HGMP)
PCP (Paracel, commercial)
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
A benign clustering situation
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
In the absence of 5‘-3‘ links
!
Two clusters corresponding to one gene
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
!
Overlap too short
Three clusters corresponding to one gene
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
!
Chimeric ESTs
!
One cluster corresponding to two genes
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Chimeric ESTs ... continued
Chimeric ESTs are quite common
Chimeric ESTs are a major nuisance for array probe selection
One of the fusion partners is usually a highly expressed mRNA
Double-picking of chimeric ESTs can fool even cautious
clustering programs.
Unigene contains several chimeric clusters
The annotation of chimeric clusters is erratic
Chimeric ESTs can be detected by genome comparison
There is one particularly bad class of chimeric sequences that
will be subject of the exercises.
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
How to select a cluster representative
If possible, pick a clone with completely known sequence
Avoid problematic regions
Alu-repeats, B1, B2 and other SINEs
LINEs
Endogenous retroviruses
Microsatellite repeats
Avoid regions with high similarity to non-identical sequences
In many clusters, orientation and position relative to ORF are
unknown and cannot be selected for.
Test selected clone for sequence correctness
Test selected clone for chimerism
Some commercial providers offer sequence verified UNIGENE
cluster representatives
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Selection of genes
If possible, use all of them
Biased selection
Selection by tissue
Selection by topic
Selection by visibility
Selection by known expression properties
Selection from unbiased pre-screen
Use sources of expression information
EST frequency
Published array studies
SAGE data
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Selection of gene regions
3‘ UTR
ORF
5‘ UTR
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Alternative polyadenylation
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Alternative polyadenylation
Constitutive polyA heterogeneity
Regulated polyA heterogeneity
3’-Fragments: reduced sensitivity
no impact on expression ratio
Fragment choice influences expression ratio
Multiple fragments necessary
Detection of cryptic polyA signals
Prediction (AATAAA)
Polyadenylated ESTs
SAGE tags
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Alternative splicing
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Alternative splicing
Constitutive splice form heterogeneity
Regulated splice form heterogeneity
Fragment in alternative exon: reduced sensitivity
No impact on expression ratio
Fragment choice influences expression ratio
Multiple fragments necessary
Detection of alternative splicing events
Hard/Impossible to predict
EST analysis (beware of pre-mRNA)
Literature
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Alternative promoter usage
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Alternative promoter usage
What is the desired readout?
If promoter activity matters most: multiple fragments
If overall mRNA level matters most: downstream fragment
Detection of alternative promoter usage
Prediction difficult (possible?)
EST analysis
Literature
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
UDP-Glucuronosyltransferases
UGT1A8
UGT1A7
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
Selection of gene regions
Coding region (ORF)
Annotation relatively safe
No problems with alternative
polyA sites
No repetitive elements or other
funny sequences
danger of close isoforms
danger of alternative splicing
might be missing in short RT
products
3’ untranslated region
Annotation less safe
danger of alternative polyA sites
danger of repetitive elements
less likely to cross-hybridize with
isoforms
little danger of alternative splicing
5’ untranslated region
close linkage to promoter
frequently not available
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11
A checklist
Pick a gene
Try get a complete cDNA sequence
Verify sequence architecture (e.g. cross-species comparison)
Mask repetitive elements (and vector!)
If possible, discard 3’-UTR beyond first polyA signal
Look for alternative splice events
Use remaining region of interest for similarity searches
Mask regions that could cross-hybridize
Use the remaining region for probe amplification or EST selection
When working with ESTs, use sequence-verified clones
Swiss Institute of Bioinformatics
Institut Suisse de Bioinformatique
LF-2001.11