|
Principle
|
Amount of hybridisation detected is proportional to no. of fragments in sample
|
|
Microarray
|
DNA probes bound to glass slide, to which sample DNA fragments can be hybridised
|
|
Probes
|
Oligonucleotides, ink-jet printed onto slides (Agilent) or synthesised in-situ (Affymetrix)
|
|
Sample
|
labelled ssDNA or antisense RNA
|
|
Technical replicate
|
Repeated measurements or procedures using the same biological sample to evaluate precision and reproducibility
|
|
Biological replicate
|
Use of multiple independent biological samples to account for biological variability
|
|
One Color Array
|
Each sample loaded into a separate microarray
|
|
Two Color Array
|
Two samples, labelled differently, loaded onto same microarray in same amounts.
Competitive hybridisation. |
|
Background Correction
|
- Adjust for non-specific hybridisation
- mismatch probes (Affymetrix) - exogenous negative control spots - remove features with lower intensity than background |
|
Log Transformation
|
Improves data distribution for classical statistical analysis
|
|
Normalisation
|
Removes systematic effects due to technical differences, which aren't due to biological differences
|
|
WithinArray Normalisation
|
Two-color microarrays;
align two channels for each array |
|
BetweenArray Normalisation
|
One-color microarrays;
Single channel platforms; Quantile normalisation, Cyclic Loess |
|
Sources of Bias
|
Dye bias, Array bias, Spatial bias
|
|
Dye Swap Design
|
pair of samples compared twice with reversed dye
|
|
Reference Design
|
Each sample hybridised against a common reference sample
|
|
Sources of Bias
|
- Differential dye incorporation
- Diff. emission response to excitation - Non-uniform focusing across the array |
|
Correction of diff responses of Cy3 and Cy5 channels
|
1. LR of Cy3 vs Cy5 intensity
2. LR of log ratio against avg. intensity (MA plot) 3. Non-linear (Loess) regression of log ratio against avg. intensity |
|
Vertical axis
|
Log ratio (R/G) = M i.e. log fold change;
M = log2 (condA exp / condB exp) |
|
Horiontal axis
|
average log-intensity between two cond;
mean(condA exp + condB exp) |
|
Non-zero intercept
|
One channel is consistently brighter than the other
|
|
Slope not equal to 1
|
One channel responds more strongly at high intensities than other
|
|
Slope not straight line
|
Non-linear relation b/w intensities of two channels
|
|
Working
|
- local reg in overlapping windows of data
- join the regression to form a smooth curve |
|
Types
|
Scaling, centering, distribution normalisation
|
|
Scaling
|
Ensure mean/median of all distributions are equal; Subtract overall mean log intensity from each log intensity
|
|
Centering
|
Ensure that all the distribution’s mean and SD are equal
|
|
Quantile normalisation
|
- sort each array in order
- average across rows - sort avg values in original order |
|
Explanatory variables
|
Covariates (quantitative measurements) and factors (categorical variables)
|
|
Levels
|
Unique values within a factor
|
|
Design or Model matrix
|
Describes the experimental design of the microarray experiment
|
|
Contrast matrix
|
Defines specific comparisons of interest b/w different experimental conditions i.e. defines the hypotheses to be tested
|
|
Types of design matrices
|
1. Mean Reference Model (with intercept)
2. Means model (wihout intercept) |
|
Means model
|
Mean gene expression levels compared independently for each sample group
|
|
Mean reference model
|
One sample group set as baseline or reference; gene exp in other groups compared relative to reference
|
|
Applications
|
- find expressed genes in a sample/cond
- find diff exp across samples/cond - exp signatures of sample/cond - exp signature of set of genes |
|
Limitations
|
- rely on prev knowledge of genome seq
- high BG levels due to cross-hyb - complex normalisations needed - limited range of detection due to BG and saturation signals |
|
Not good for
|
- determining exp or diff exp of single or small set of genes
- accurately studying protein levels and functional activity (measure steady-state levels of RNA transcripts) - absolute level of exp of a gene (true conc of mRNA) |
No comments yet.
Cheatography n. The art of mapping complex subjects into clear, instantly navigable reference guides.
