- Page 1: Probe-Level Analysis ofAffymetrix G
- Page 6: DNA• A deoxyribonucleic acid or D
- Page 9 and 10: Differential expression• Each cel
- Page 11 and 12: Central dogma• The expression of
- Page 13 and 14: Idea: measure the amount of mRNA to
- Page 15 and 16: Probes and Probesets
- Page 17 and 18: Constructing the Chip
- Page 19 and 20: Hybridization to the Chip
- Page 21 and 22: Chip dat file - checkered board - c
- Page 23 and 24: Levels of Analysis• Probe level-
- Page 25 and 26: Probe-Level Analysis• What is Pro
- Page 27 and 28: RMA Background Approach• Convolut
- Page 29 and 30: Normalization“Non-biological fact
- Page 31 and 32: Non-linear normalizationneededUnnor
- Page 33 and 34: Sort Sort columns of of originalmat
- Page 35 and 36: General Probe Level Modelyij= f( X)
- Page 37 and 38: Probe Pattern SuggestsIncluding Pro
- Page 39 and 40: Summarization• Problem: Calculati
- Page 41 and 42: RMA mostly does well inpracticeDete
- Page 43 and 44: The RMA modely = m+ α + β + εij
- Page 45 and 46: • Advantages-Fast- Very robust•
- Page 47 and 48: Variance CovarianceEstimates• Sup
- Page 49 and 50: Detecting DifferentialExpression•
- Page 51 and 52: Affymetrix Spike-in Data• 59 chip
- Page 53 and 54:
Fold ChangeFC = X − XlmWhereXl=
- Page 55 and 56:
“Robust” t-statistict=X% − X%
- Page 57 and 58:
Limma “ebayes” t-statistic• G
- Page 59 and 60:
Probe Level Model teststatisticstPL
- Page 65:
What Happens as the Numberof Arrays
- Page 71 and 72:
ResultsMethod Individual Models Sin
- Page 75 and 76:
What is Going On Here?• Examine r
- Page 77 and 78:
Low Non-Differential
- Page 79 and 80:
High Non-Differential
- Page 81 and 82:
GeneLogic Tonsil dataset
- Page 83 and 84:
Mixture Data ResultsMethod 3 vs 3 4
- Page 85 and 86:
Quality Assessment using PLM• PLM
- Page 87 and 88:
An Image Gallery“Crop Circles”
- Page 89 and 90:
RLE PlotsRelativeLogExpression
- Page 91 and 92:
Discordant Arrays
- Page 93 and 94:
Ongoing Work in this Area• Techno
- Page 95 and 96:
Additional Slides
- Page 97 and 98:
Chip dat file - checkered board - o
- Page 99 and 100:
From Chip To Data
- Page 101 and 102:
Computing ExpressionMeasures:A Thre
- Page 103 and 104:
Background Signal Methods• Affyme
- Page 105 and 106:
Convolution Model• O = S + N- O i
- Page 107 and 108:
What About Non Spike-ins?• We don
- Page 109 and 110:
Establishing a RelationshipγBetwee
- Page 111 and 112:
Normalization Methods• Methods al
- Page 113 and 114:
RMA Model• To each probeset (k),
- Page 115 and 116:
Advantages/Disadvantages of RMA/Med
- Page 117 and 118:
Comparing the backgroundmethods•
- Page 119 and 120:
Assessing Bias: ObservedFold-change
- Page 127 and 128:
Detecting DifferentialExpression: R
- Page 129:
Comparing theNormalization Methods
- Page 134 and 135:
Variability of Non-DifferentialGene
- Page 136 and 137:
ROC Curves
- Page 138 and 139:
Slope ValueAll 0.493Mid 0.665Low 0.
- Page 140 and 141:
Slope ValueAll 0.589Mid 0.751Low 0.
- Page 142 and 143:
Slope ValueAll 0.695Mid 0.82Low 0.5
- Page 144 and 145:
Slope: 0.484
- Page 146 and 147:
Slope: 0.583
- Page 148 and 149:
Slope: 0.692
- Page 150:
What About the TreatmentEffect Mode