Contents

Author: Martin Morgan
Date: 22 July, 2019

1 R

1.1 History of R and CRAN

  • Statistical programming language. Concieved 1992, initial version 1996, stable beta version in 2000; an implementation of S. CRAN started in 1997.
  • ‘Free’ software: no cost, open source, broad use.
  • Extensible: packages (15,000 on CRAN, 1750 on Bioconductor)
  • Key features
    • Intrinsic statistical concepts
    • Vectorized computation
    • ‘Old-school’ scripts rather than graphical user interface – great for reproducibility!
    • (Advanced) copy-on-change semanatics

1.2 Vectors and data frames

1 + 2
## [1] 3
x = c(1, 2, 3)
1:3             # sequence of integers from 1 to 3
## [1] 1 2 3
x + c(4, 5, 6)  # vectorized
## [1] 5 7 9
x + 4           # recycling
## [1] 5 6 7

Vectors

  • numeric(), character(), logical(), integer(), complex(), …
  • NA: ‘not available’
  • factor(): values from restricted set of ‘levels’.

Operations

  • numeric: ==, <, <=, >, >=, …
  • logical: | (or), & (and), ! (not)
  • subset: [, e.g., x[c(2, 3)]
  • assignment: [<-, e.g., x[c(1, 3)] = x[c(1, 3)]
  • other: is.na()

Functions

x = rnorm(100)
y = x + rnorm(100)
plot(x, y)

  • Many!

data.frame

df <- data.frame(Independent = x, Dependent = y)
head(df)
##   Independent  Dependent
## 1  -0.4338047 -0.5779168
## 2  -0.2769985 -1.0665115
## 3  -1.6966211 -1.8769578
## 4  -0.6481076 -0.9540841
## 5  -2.1015776 -1.1166887
## 6   0.7109163 -0.3363154
df[1:5, 1:2]
##   Independent  Dependent
## 1  -0.4338047 -0.5779168
## 2  -0.2769985 -1.0665115
## 3  -1.6966211 -1.8769578
## 4  -0.6481076 -0.9540841
## 5  -2.1015776 -1.1166887
df[1:5, ]
##   Independent  Dependent
## 1  -0.4338047 -0.5779168
## 2  -0.2769985 -1.0665115
## 3  -1.6966211 -1.8769578
## 4  -0.6481076 -0.9540841
## 5  -2.1015776 -1.1166887
plot(Dependent ~ Independent, df)  # 'formula' interface

  • List of equal-length vectors
  • Vectors can be of different type
  • Two-dimensional subset and assignment
  • Column access: df[, 1], df[, "Indep"], df[[1]], df[["Indep"]], df$Indep

Exercise: plot only values with Dependent > 0, Independent > 0

  1. Select rows

    ridx <- (df$Dependent > 0) & (df$Independent > 0)

  2. Plot subset

    plot(Dependent ~ Independent, df[ridx, ])

  3. Skin the cat another way

    plot(
    Dependent ~ Independent, df,
    subset = (Dependent > 0) & (Independent > 0)
)

1.3 Analysis: functions, classes, methods

fit <- lm(Dependent ~ Independent, df)  # linear model -- regression
anova(fit)                              # summary table
## Analysis of Variance Table
## 
## Response: Dependent
##             Df  Sum Sq Mean Sq F value    Pr(>F)    
## Independent  1  92.664  92.664   70.32 3.787e-13 ***
## Residuals   98 129.139   1.318                      
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
plot(Dependent ~ Independent, df)
abline(fit)

  • lm(): plain-old function
  • fit: an object of class “lm”
  • anova(): a generic with a specific method for class “lm”
class(fit)
## [1] "lm"
methods(class="lm")
##  [1] add1           alias          anova          case.names    
##  [5] coerce         confint        cooks.distance deviance      
##  [9] dfbeta         dfbetas        drop1          dummy.coef    
## [13] effects        extractAIC     family         formula       
## [17] hatvalues      influence      initialize     kappa         
## [21] labels         logLik         model.frame    model.matrix  
## [25] nobs           plot           predict        print         
## [29] proj           qr             residuals      rstandard     
## [33] rstudent       show           simulate       slotsFromS3   
## [37] summary        variable.names vcov          
## see '?methods' for accessing help and source code

1.4 Help!

?"plot"          # plain-old-function or generic
?"plot.formula"  # method
?"plot.lm"       # method for object of class 'lm', plot(fit)

1.5 Packages

library(ggplot2)
ggplot(df, aes(x = Independent, y = Dependent)) +
    geom_point() + geom_smooth(method = "lm")

  • General purpose: >15,000 packages on CRAN
  • Gain contributor’s domain expertise and weird (or other) idiosyncracies
  • Installation (once only per computer) versus load (via library(ggplot2), once per session)

2 Bioconductor

Started 2002 as a platform for understanding analysis of microarray data

2.1 Packages

1,750 packages. Domains of expertise:

  • Sequencing (RNASeq, ChIPSeq, single-cell, called variants, …)
  • Microarrays (methylation, expression, copy number, …)
  • flow cytometry
  • proteomics
  • …

Important themes

  • Reproducible research
  • Interoperability between packages & work kflows
  • Usability

Resources

2.2 Objects

A distinctive feature of Bioconductor – use of objects for representing data

library(Biostrings)
dna <- DNAStringSet(c("AACTCC", "CTGCA"))
dna
##   A DNAStringSet instance of length 2
##     width seq
## [1]     6 AACTCC
## [2]     5 CTGCA
reverseComplement(dna)
##   A DNAStringSet instance of length 2
##     width seq
## [1]     6 GGAGTT
## [2]     5 TGCAG
  • Biostrings: DNA, RNA, AA representation and manipulation
  • GenomicRanges: Coordinates in genome space
  • SummarizedExperiment: coordinating ‘assay’ data (e.g., counts from an RNASeq experiment) with row and column annotations (e.g., information about samples and experimental treatments).

2.3 High-throughput sequence work flow