bch441-work-abc-units/BIN-Sequence.R

# BIN-Sequence.R
#
# Purpose:  A Bioinformatics Course:
#              R code accompanying the BIN-Sequence unit.
#
# Version:  1.4
#
# Date:     2017  09  - 2019  01
# Author:   Boris Steipe (boris.steipe@utoronto.ca)
#
# Versions:
#           1.4    Change from require() to requireNamespace(),
#                      use <package>::<function>() idiom throughout,
#                      use Biocmanager:: not biocLite()
#           1.3    Update set.seed() usage
#           1.2    Removed irrelevant task. How did that even get in there? smh
#           1.1    Add chartr()
#           1.0    First live version 2017.
#
# TODO:
#
#
# == DO NOT SIMPLY  source()  THIS FILE! =======================================
#
# If there are portions you don't understand, use R's help system, Google for an
# answer, or ask your instructor. Don't continue if you don't understand what's
# going on. That's not how it works ...
#
# ==============================================================================


#TOC> ==========================================================================
#TOC> 
#TOC>   Section  Title                                Line
#TOC> ----------------------------------------------------
#TOC>   1        Prepare                                63
#TOC>   2        Storing Sequence                       80
#TOC>   3        String properties                     109
#TOC>   4        Substrings                            116
#TOC>   5        Creating strings: sprintf()           137
#TOC>   6        Changing strings                      172
#TOC>   6.1.1          Changing case                   174
#TOC>   6.1.2          Reverse                         179
#TOC>   6.1.3          Change characters               183
#TOC>   6.1.4          Substitute characters           211
#TOC>   6.2        stringi and stringr                 231
#TOC>   6.3        dbSanitizeSequence()                241
#TOC>   7        Permuting and sampling                253
#TOC>   7.1        Permutations                        260
#TOC>   7.2        Sampling                            306
#TOC>   7.2.1          Equiprobable characters         308
#TOC>   7.2.2          Defined probability vector      350
#TOC> 
#TOC> ==========================================================================


#
#
#
#


# =    1  Prepare  =============================================================

# Much basic sequence handling is supported by the Bioconductor package
# Biostrings.

if (! requireNamespace("BiocManager", quietly = TRUE)) {
  install.packages("BiocManager")
}
if (! requireNamespace("Biostrings", quietly = TRUE)) {
  BiocManager::install("Biostrings")
}
# Package information:
#  library(help = Biostrings)       # basic information
#  browseVignettes("Biostrings")    # available vignettes
#  data(package = "Biostrings")     # available datasets


# =    2  Storing Sequence  ====================================================


# Sequences can be represented and stored as vectors of single characters ...
(v <- c("D", "I", "V", "M", "T", "Q"))

# ... as strings ...
(s <- "DIVMTQ")

# ... or as more complex objects with rich metadata e.g. as a Biostrings
# DNAstring, RNAstring, AAString, etc.
(a <- Biostrings::AAString("DIVMTQ"))

# ... and all of these representations can be interconverted:

# string to vector ...
unlist(strsplit(s, ""))

# vector to string ...
paste(v, sep = "", collapse = "")

# ... and AAstring to plain string.
as.character(a)

# Since operations with character vectors trivially follow all other vector
# conventions and syntax, and we will look at Biostrings methods in more
# detail in a later unit, we will focus on basic strings in the following.


# =    3  String properties  ===================================================


length(s) # why ???
nchar(s)  # aha


# =    4  Substrings  ==========================================================

# Use the substr() function
substr(s, 2, 4)

# or the similar substring()
substring(s, 2, 4)

# Note: both functions are vectorized (i.e. they operate on vectors
# of arguments, you don't need to loop over input)...
myBiCodes <- c("HOMSA", "MUSMU", "FUGRU", "XENLA")
substr(   myBiCodes, 1, 3)
substring(myBiCodes, 1, 3)

# ... however only substring() will also use vectors for start and stop
s <- "gatattgtgatgacccagtaa"     # a DNA sequence
(i <- seq(1, nchar(s), by = 3))  # an index vector
substr(   s, i, i+2)             # ... returns only the first nucleotide triplet
substring(s, i, i+2)             # ... returns all triplets


# =    5  Creating strings: sprintf()  =========================================


# Sprintf is a very smart, very powerful function and has cognates in all
# other programming languages. It has a bit of a  learning curve, but this is
# totally worth it:
# the function takes a format string, and a list of other arguments. It returns
# a formatted string. Here are some examples - watch carefully for sprintf()
# calls elsewhere in the code.

sprintf("Just a string.")
sprintf("A string and the number %d.", 5)
sprintf("More numbers: %d ate %d.", 7, 9) # Sorry
sprintf("Pi is ~ %1.2f ...", pi)
sprintf("or more accurately ~ %1.11f.", pi)
x <- "bottles of beer"
N <- 99
sprintf("%d %s on the wall, %d %s - \ntake %s: %d %s on the wall.",
        N, x, N, x, "one down, and pass it around", N - 1, x)

# Note that in the last example, the value of the string was displayed with
# R's usual print-formatting function and therefore the line-break "\n" did
# not actually break the line. To have line breaks, tabs etc, you need to use
# cat() to display the string:

for (i in N:(N-4)) {
  cat(sprintf("%d %s on the wall, %d %s - \ntake %s: %d %s on the wall.\n\n",
              i, x, i, x, "one down, and pass it around", i - 1, x))
}

# sprintf() is vectorized: if one of its parameters is a vector, it
# will generate one output string for each of the vector's elements:
cat(sprintf("\n%s fish", c("one", "two", "red", "blue")))


# =    6  Changing strings  ====================================================

# ===   6.1.1  Changing case              
tolower(s)
toupper(tolower(s))


# ===   6.1.2  Reverse                    
reverse(s)


# ===   6.1.3  Change characters          
# chartr(old, new, x) maps all characters in x that appear in "old" to the
# correpsonding character in "new."

chartr("aeio", "uuuu", "We hold these truths to be self-evident ...")

# One could implement toupper() and tolower() with this - remember that R has
# character vectors of uppercase and lowercase letters as language constants.
chartr(paste0(letters, collapse = ""),
       paste0(LETTERS, collapse = ""),
       "Twinkle, twinkle little star, how I wonder what you are.")

# One amusing way to use the function  is for a reversible substitution
# cypher.
set.seed(112358)                       # set RNG seed for repeatable randomness
(myCypher <- paste0(sample(letters), collapse = ""))
set.seed(NULL)                         # reset the RNG

(lett <- paste0(letters, collapse = ""))

# encode ...
(x <- chartr(lett, myCypher, "... seven for a secret, never to be told."))

# decode ...
chartr(myCypher, lett, x)
# (Nb. substitution cyphers are easy to crack!)


# ===   6.1.4  Substitute characters      
(s <- gsub("IV", "i-v", s))  # gsub can change length, first argument is
                             # a "regular expression"!

# I use it often to delete characters I don't want ...
# ... select something, and substitute the empty string for it.
(s <- gsub("-", "", s))

# For example: clean up a sequence
# copy/paste from UniProt
(s <- "        10         20         30         40         50
MSNQIYSARY SGVDVYEFIH STGSIMKRKK DDWVNATHIL KAANFAKAKR ")


# remove numbers
(s <- gsub("[0-9]", "", s))

# remove "whitespace" (spaces, tabs, line breaks)...
(s <- gsub("\\s", "", s))

# ==   6.2  stringi and stringr  ===============================================

# But there are also specialized functions eg. to remove leading/trailing
# whitespace which may be important to sanitize user input etc. Have a look at
# the function descriptions for the stringr and the stringi package. stringr is
# part of the tidyverse, and for the most part a wrapper for stringi functions.
# https://github.com/tidyverse/stringr



# ==   6.3  dbSanitizeSequence()  ==============================================

# In our learning units, we use a function dbSanitizeSequence() to clean up
# sequences that may be copy/pasted from Web-sources

s <- ">FASTA header will be removed
10         20         30         40         50
MSNQIYSARY SGVDVYEFIH STGSIMKRKK DDWVNATHIL KAANFAKAKR "

dbSanitizeSequence(s)


# =    7  Permuting and sampling  ==============================================


# An important aspect of working with strings is generating random strings
# with given statistical properties: reference items to evaluate significance.


# ==   7.1  Permutations  ======================================================


# One way to produce such reference items is to permute a string. A permuted
# string has the same composition as the original, but all positional
# information is lost. The sample() function can be used to permute:

# This is the sequence of the ompA secretion signal
(s <- unlist(strsplit("MKKTAIAVALAGFATVAQA", "")))

(x <- sample(s, length(s)))  # permuted

# Here's a small example how such permuted strings may be useful. As you look
# at the ompA sequence, you suspect that the two lysines near the +-charged
# N-terminus may not be accidental, but selected for a positively charged
# N-terminus. What is the chance that such a sequence has two lysines close to
# the N-terminus simply by chance? Or put differently: what is the average
# distance of two lysines in such a sequence to the N-terminus. First, we
# need an expression that measures the distance. A simple use of the which()
# function will do just fine.

which(s == "K")        # shows they are in position 2 and 3, so ...
mean(which(s == "K"))  # ... gives us the average, and ...
mean(which(x == "K"))  # ... gives us the average of the permuted sequence.

# So what does the distribution look like? Lets do 10,000 trials.

(s <- unlist(strsplit("MKKTAIAVALAGFATVAQA", "")))
N <- 10000
d <- numeric(N)

set.seed(112358)                       # set RNG seed for repeatable randomness
for (i in 1:N) {
  d[i] <- mean(which(sample(s, length(s)) == "K"))
}
set.seed(NULL)                         # reset the RNG

hist(d, breaks = 20)
abline(v = 2.5, lwd = 2, col = "firebrick")
sum(d <= 2.5) # 276. 276 of our 10000 samples are just as bunched near the
              # N-terminus or more. That's just below the signifcance
              # threshold of 5 %. It's a trend, but to be sure we are looking
              # at a biological effect we would need to see more
              # sequences.


# ==   7.2  Sampling  ==========================================================

# ===   7.2.1  Equiprobable characters    

# Assume you need a large random-nucleotide string for some statistical model.
# How to create such a string? sample() can easily create it:

nuc <- c("A", "C", "G", "T")
N <- 100

set.seed(16818)                        # set RNG seed for repeatable randomness
v <- sample(nuc, N, replace = TRUE)
set.seed(NULL)                         # reset the RNG

(mySeq <- paste(v, collapse = ""))

# What's the GC content?
table(v)
sum(table(v)[c("G", "C")]) # 51 is close to expected

# What's the number of CpG motifs? Easy to check with the stringi
# stri_match_all() function

if (! requireNamespace("stringi", quietly = TRUE)) {
  install.packages("stringi")
}
# Package information:
#  library(help = stringi)       # basic information
#  browseVignettes("stringi")    # available vignettes
#  data(package = "stringi")     # available datasets


(x <- stri::stri_match_all(mySeq, regex = "CG"))
length(unlist(x))

# Now you could compare that number with yeast DNA sequences, and determine
# whether there are more or less CpG motifs than expected by chance.
# (cf. https://en.wikipedia.org/wiki/CpG_site)
# But hold on: is that a fair comparison? sample() gives us all four nucleotides
# with the same probability. But the yeast genomic DNA GC content is only
# 38%. So you would expect fewer CpG motifs based on the statistical properties
# of the smaller number of Cs and Gs - before biology even comes into play. How
# do we account for that?

# ===   7.2.2  Defined probability vector 

# This is where we need to know how to create samples with specific probability
# distributions. A crude hack would be to create a sampling source vector with
# 19 C, 19 G, 31 A and 31 T
c(rep("C", 19), rep("G", 19), rep(c("A"), 31), rep(c("T"), 31))
# ... but that doesn't scale if the numeric accuracy needs to be higher.
#
# However sample() has an argument that takes care of that: you can explicitly
# specify the probabilities with which each element of the the sampling vector
# should be chosen:

nuc <- c("A", "C", "G", "T")
N <- 100
myProb <- c(0.31, 0.19, 0.19, 0.31)    # sampling probabilities

set.seed(16818)                       # set RNG seed for repeatable randomness
v <- sample(nuc, N, prob = myProb, replace = TRUE)
set.seed(NULL)                         # reset the RNG

(mySeq <- paste(v, collapse = ""))

# What's the GC content?
table(v)
sum(table(v)[c("G", "C")]) # Close to expected

# What's the number of CpG motifs?
(x <- stringi::stri_match_all(mySeq, regex = "CG"))
# ... not a single one in this case.



# [END]