Chapter 4 Lab 3: Logical Subsetting and Conditional Filtering

Objectives:

  1. To understand Boolean logic (AND, OR, NOT) in programming
  2. To use logical comparisons to build TRUE/FALSE vectors
  3. To use conditional filters to isolate specific rows of a data set
  4. To apply subsetting to isolate biological variables or categories of interest

Last lab we learned to build vectors and data frames. Today we will learn how to ask R questions about our data and pull out only the rows that answer “yes.”

4.1 Logical comparisons

R can evaluate whether a statement is TRUE or FALSE using comparison operators:

Operator Meaning
== equal to
!= not equal to
> greater than
< less than
>= greater than or equal to
<= less than or equal to 
5 > 3
## [1] TRUE
5 == 3
## [1] FALSE
"cat" == "dog"
## [1] FALSE
"cat" == "cat"
## [1] TRUE

These comparisons also work element-wise on a whole vector:

nums <- c(2, 5, 8, 1, 9, 4)
nums > 4
## [1] FALSE  TRUE  TRUE FALSE  TRUE FALSE

Question 1

  • Create a numeric vector called ages with 8 numbers between 1 and 100
  • Write the code to check which elements of ages are greater than 30
  • Write the code to check which elements of ages are exactly equal to 18

4.2 Boolean logic: AND, OR, NOT

Often we want to combine more than one condition:

Symbol Meaning
& AND — both conditions must be TRUE
\| OR — at least one condition must be TRUE
! NOT — flips TRUE to FALSE and vice versa 
x <- c(2, 5, 8, 1, 9, 4)

# AND: greater than 3 AND less than 9
x > 3 & x < 9
## [1] FALSE  TRUE  TRUE FALSE FALSE  TRUE
# OR: less than 2 OR greater than 8
x < 2 | x > 8
## [1] FALSE FALSE FALSE  TRUE  TRUE FALSE
# NOT: the reverse of x > 4
!(x > 4)
## [1]  TRUE FALSE FALSE  TRUE FALSE  TRUE

Question 2

  • Using your ages vector from Question 1, write code that identifies which values are between 18 and 65 (inclusive) using &
  • Write code that identifies which values are younger than 18 OR older than 65 using |
  • Explain, in your own words, the difference between & and |

4.3 Subsetting vectors and data frames

Once you have a logical (TRUE/FALSE) vector, you can use it inside square brackets [ ] to pull out only the elements where the condition is TRUE.

x <- c(2, 5, 8, 1, 9, 4)
x[x > 4]
## [1] 5 8 9

The same idea extends to data frames, but now we have to say whether we are filtering rows or columns. Recall from Lab 2 that data frames use dataframe[rows, columns] syntax.

species <- c("frog", "frog", "toad", "toad", "salamander")
length_mm <- c(45, 52, 38, 41, 60)
site <- c("pond1", "pond2", "pond1", "pond2", "pond1")

herps <- data.frame(species, length_mm, site, stringsAsFactors = FALSE)
herps
##      species length_mm  site
## 1       frog        45 pond1
## 2       frog        52 pond2
## 3       toad        38 pond1
## 4       toad        41 pond2
## 5 salamander        60 pond1

To get only the rows where species is "frog":

herps[herps$species == "frog", ]
##   species length_mm  site
## 1    frog        45 pond1
## 2    frog        52 pond2

The subset() function does the same thing with slightly friendlier syntax:

subset(herps, species == "frog")
##   species length_mm  site
## 1    frog        45 pond1
## 2    frog        52 pond2

You can combine conditions with & and | inside subset() as well:

subset(herps, species == "frog" & length_mm > 48)
##   species length_mm  site
## 2    frog        52 pond2

Question 3

  • Using the herps data frame above (or a similar data frame you build with at least 8 rows and one categorical + one numeric column), write code to:
    • Subset only the rows collected at “pond1”
    • Subset only the rows where length_mm is greater than 40 AND site is “pond1”
    • Subset the rows where species is NOT “toad” (use !=)
  • For each, explain in one sentence what biological question that subset could help answer

4.4 Isolating specific variants or categories of interest

In genomics and ecology, filtering is often used to isolate specific variants, genes, or categories of interest from a much larger data set before further analysis.

For example, to make plots like this:

We need to have these types of data:

gene_id <- c("g001","g002","g003","g004","g005","g006")
chromosome <- c(1, 1, 2, 3, 3, 2)
expression <- c(12.4, 0.3, 8.9, 15.2, 0.1, 22.7)

gene.table <- data.frame(gene_id, chromosome, expression, stringsAsFactors = FALSE)
gene.table
##   gene_id chromosome expression
## 1    g001          1       12.4
## 2    g002          1        0.3
## 3    g003          2        8.9
## 4    g004          3       15.2
## 5    g005          3        0.1
## 6    g006          2       22.7

For example, to isolate genes on chromosome 3 that show high expression (say, above 10):

subset(gene.table, chromosome == 3 & expression > 10)
##   gene_id chromosome expression
## 4    g004          3       15.2

Question 4

  • Using the gene.table example (or a similar table of your own with at least 8 rows), write code to isolate genes with expression below 1 (candidates for “not expressed”)
  • Write code to isolate genes on chromosome 1 OR chromosome 2
  • Why might a biologist want to filter out low-expression genes before doing further statistical analysis? (This is a conceptual question — think about noise vs. signal)