Practical demonstration of R and Bioconductor

1
Practical demonstration of R and Bioconductor

• Elva Diaz
• University of California, Davis
• Department of Med-Pharmacology
• June 19, 2005

2
What is R?

• R is a language and environment for statistical
  computing and graphics
• Open source implementation of S language
• Provides variety of statistical and graphical
  techniques implemented as packages
• marray and limma are R packages
• R is available as free software from
• http//www.r-project.org/

3
Introduction to linear models

• Define M and A values
• M log2(R/G)
• A ½(log2R log2G)
• Use linear models to estimate M (referred to as
  fitted M values or M)
• Simple linear models
• Two replicates M ½(M1 M2)
• Dye swap M ½ (M1 - M2)

4
Complex Linear Models

• Example with two samples (A, B) and common
  reference (CR)
• M1 A/CR
• M2 B/CR
• Use linear model to estimate A/B

A
B
CR
5
Design Matrix

• The columns represent the parameters to be
  estimated by the linear model
• The rows are the slides in the experiments
• Parameterization 1 Parameterization 2

A/CR B/CR
M1 1 0
M2 0 1
A/CR A/B
M1 1 0
M2 1 -1
6
Example dataset

• We will use a subset of the weaver dataset
  described in Diaz et al, 2002.

7
Cerebellum development
Math-1
Cyclin D2
L1
GABA-AR a6
8
Weaver mice

• Mutation in GIRK2 K channel gene
• Lack mature granule cells
• Quantify difference in expression profile
  compared to wild type

9
Experimental design

• Two time points (P11, P21)
• Two mouse strains (WT, MT)
• Eight hybridizations (arrows)

WT P11
WT P21
CREF
MT P11
MT P21
Cy5 ? Cy3
10
Linear model parameters

• Five different samples
• P11WT, P11MT, P21WT, P21MT, CREF
• Estimate compared with CREF
• P11WT-CREF
• P11MT-CREF
• P21WT-CREF
• P21MT-CREF

11
Linear model
P11WT -CREF 1 0 0 0 1 0 0 -1
P21WT -CREF 0 0 0 1 0 0 -1 1
P11MT -CREF 0 1 0 0 -1 1 0 0
P11MT -CREF 0 0 1 0 0 -1 1 0
Slide Exp 1 Exp 2 Exp 3 Exp 4 Exp 5 Exp 6 Exp
7 Exp 8
Cy5 P11WT P11MT P21MT P21WT P11WT P11MT P21MT P21W
T
Cy3 CREF CREF CREF CREF P11MT P21MT P21WT P11WT
12
Contrasts

• Pairwise combination of linear model output to
  make all comparisons
• P11WT-P21WT
• P21WT-P21MT
• P21MT-P11MT
• P11MT-P11WT

13
Differentially expressed genes

• Identify differentially expressed genes at P21
  between wild type and mutant
• That is, rank genes by the contrast P21MT
  P21WT
• Expression of the GABA-AR a6 subunit is absent in
  weaver (large, negative P21MT P21WT value)

14
marray input files

• Hybridizations file (WeaverSample.txt)
• GAL file (Weaver.gal)
• Image analysis files (filename.gpr)
• All are tab-delimited text files

15
Switch to R tutorial

• See handout R_Bioconductor_CSHL2005