Signal Detection, Estimate of Fold Change

1
Signal Detection,Estimate of Fold Change
Differential Expressionin Affymetrix GeneChips
2

• Each GeneChip measures 12 000 genes, each gene
  represented by 11 to 20 PM and MM oligo
  probes scattered across chip.
• Intensity data from each chip is stored in a .cel
  file containing an intensity value at each site
• Bioconductor MAS software uses information in
  .cdf file to interpret this as an intensity value
  for each PM and each MM probe.

3

• Given a set of typically 16 PM and MM intensity
  values (number of replicate chips in expt.), how
  can we obtain a measure of mRNA expression for a
  given gene?
• Either as an absolute mRNA concentration
• Or a relative change in mRNA concentration
  between treatments

4

• (Absolute concentration) no easy answer
• (Relative expression between treatments)
  Expression measures such as
• MAS5
• RMA
• Li-Wong
• can be useful.
• Bioconductor provides inbuilt functions for these
  measures

5
MAS5 (MicroArray Suite v5)

• MM subtraction

if
where
something lt PM otherwise
2. Tukey biweight average of logged Vs within
probeset (summarisation)
SignalLogValue
6

• 3. Optional scaling factor

4. Final output is
Reported value of ith probeset
7
RMA (Robust Microarray Average)
Irizarry et al. Biostatistics, 4 (2003) 249-264
1. Background Correction
Subtract from PMs a probe specific background
correction using a model based on observed
intensity being the sum of (exponential) signal
(normal) noise.

• 2. Quantile normalisation

Assuming multiple replicates of each experiment,
this adjusts intensities so that the
distribution of intensities is the same for all
chips within set of replicates.
8

• 3. Take logs

4. Average across the 16 probes in probeset using
median polish summarisation
i.e., fit to model
is the required measure
9
Affymetrix Latin Square experiment

• 14 genes spiked at cyclic permutations of the 14
  concentrations (0, 0.25, 0.5, 1, ,1024) pM
  
• into background of human pancreas cRNA
• Hybridised onto 14 arrays
• 3 replicates of experiment

10
GENES
CHIPS
11
Raw data from .cel files
Affy spike-in experiment Gene 37777_at Red
PM Black MM
12
a b c d e f g h i j k l m
n q
rma mas5 conc
(log scale)
13
a b c d e f g h i j k l m
n q
rma log2(mas5) conc (log
scale)
14
a b c d e f g h i j k l m
n q
rma log2(mas5) conc (log
scale)
15
Gene 37777_at
Background
64 pM
Saturation
1 pM
16
(No Transcript)
17
Lesson Expression measures underestimate fold
change!
18
a b c d e f g h i j k l m
n q
rma mas5 conc
(log scale)
19
GENES
CHIPS
20
rma mas5 conc
(log scale)
21
rma mas5 conc
(log scale)
22
GeneLogic Dilution/Mixture study

• 20 µg/200 ml solutions of liver and central
  nervous system cell line cRNA diluted to samples
  of (20, 10, 7.5, 5, 2.5, 1.25) µg/200 ml
• Hybridised onto Affymetrix HG_U95av2 chips
• 5 replicates of each dilution/mixture

23
(No Transcript)
24

• M log (liver 20µg) log (liver 1.25µg)
• A log (liver 20µg) log (liver 1.25µg)

25
M log (CNS 10µg) log (liver 10µg) A
log (CNS 10µg) log (liver 10µg)
26
Pairwise comparisons of M log (CNS 10µg)
log (liver 10µg)
27
Langmuir Adsorption Model

• Assume
• (Adsorption) Target mRNA attaches to probes at a
  rate proportional to concentration of specific
  target mRNA and fraction of unoccupied probes
• (Desorption) Target mRNA detaches from probes at
  a rate proportional to fraction of occupied
  probes
• ? At equilibrium, intensity I(x) at target
  concentration x follows Langmuir Isotherm

28
(No Transcript)
29
Raw data from .cel files
Affy spike-in experiment Gene 37777_at Red
PM Black MM
30
Raw data from .cel files
Affy spike-in experiment Gene 37777_at Red
PM Black MM
31
Raw data from .cel files
Affy spike-in experiment Gene 37777_at Red
PM Black MM
32

• Individual probes have very different responses
  depending on their nucleotide sequence!
• Temperature, pH, wafer effects, time to reach
  equilibrium etc. also important
• Role (and usefulness) of MMs is not clear
• - The problem of extracting absolute
  concentration values from .cel file data still
  not solved

33
Summary

• Expression measures such as MAS5, RMA, Li-Wong
  provide measures of relative concentrations of
  same gene under different treatments

34
Summary

• Expression measures such as MAS5, RMA, Li-Wong
  provide measures of relative concentrations of
  same gene under different treatments
• But not different genes within same treatment

35
Summary

• Expression measures such as MAS5, RMA, Li-Wong
  provide measures of relative concentrations of
  same gene under different treatments
• But not different genes within same treatment
• Fold change underestimated at high concentrations
  (saturation) or low concentrations (background
  and random noise)

36
Summary

• Expression measures such as MAS5, RMA, Li-Wong
  provide measures of relative concentrations of
  same gene under different treatments
• But not different genes within same treatment
• Fold change underestimated at high concentrations
  (saturation) or low concentrations (background
  and random noise)
• Normalisation (e.g. quantile normalisation in
  RMA) should only be used within a set of
  replicates

37
Summary

• Expression measures such as MAS5, RMA, Li-Wong
  provide measures of relative concentrations of
  same gene under different treatments
• But not different genes within same treatment
• Fold change underestimated at high concentrations
  (saturation) or low concentrations (background
  and random noise)
• Normalisation (e.g. quantile normalisation in
  RMA) should only be used within a set of
  replicates
• Scaling (e.g. MAS5) assumes same total target
  concentration between replicates/treatments

38
Summary

• Expression measures such as MAS5, RMA, Li-Wong
  provide measures of relative concentrations of
  same gene under different treatments
• But not different genes within same treatment
• Fold change underestimated at high concentrations
  (saturation) or low concentrations (background
  and random noise)
• Normalisation (e.g. quantile normalisation in
  RMA) should only be used within a set of
  replicates
• Scaling (e.g. MAS5) assumes same total target
  concentration between replicates/treatments
• There is no reliable measure of absolute target
  concentration from intensity data