Parallel Protein Sequence Similarity Calculation Based on Sequence Alignment (Disk-Based Version)

Parallel calculation of protein sequence similarity based on sequence alignment. This version offloads the partial results in each batch to the hard drive and merges the results together in the end, which reduces the memory usage.

Usage

parSeqSimDisk(
  protlist,
  cores = 2,
  batches = 1,
  path = tempdir(),
  verbose = FALSE,
  type = "local",
  submat = "BLOSUM62",
  gap.opening = 10,
  gap.extension = 4
)

Arguments

protlist

A length n list containing n protein sequences, each component of the list is a character string, storing one protein sequence. Unknown sequences should be represented as "".

cores

Integer. The number of CPU cores to use for parallel execution, default is 2. Users can use the availableCores() function in the parallelly package to see how many cores they could use.

batches

Integer. How many batches should we split the pairwise similarity computations into. This is useful when you have a large number of protein sequences, enough number of CPU cores, but not enough RAM to compute and fit all the pairwise similarities into a single batch. Defaults to 1.

path

Directory for caching the results in each batch. Defaults to the temporary directory.

verbose

Print the computation progress? Useful when batches > 1.

type

Type of alignment, default is "local", can be "global" or "local", where "global" represents Needleman-Wunsch global alignment; "local" represents Smith-Waterman local alignment.

submat

Substitution matrix, default is "BLOSUM62", can be one of "BLOSUM45", "BLOSUM50", "BLOSUM62", "BLOSUM80", "BLOSUM100", "PAM30", "PAM40", "PAM70", "PAM120", or "PAM250".

gap.opening

The cost required to open a gap of any length in the alignment. Defaults to 10.

gap.extension

The cost to extend the length of an existing gap by 1. Defaults to 4.

Value

A n x n similarity matrix.

See also

See parSeqSim for the in-memory version.

Author

Nan Xiao <https://nanx.me>

Examples

if (FALSE) { # \dontrun{

# Be careful when testing this since it involves parallelization
# and might produce unpredictable results in some environments

library("Biostrings")
library("foreach")
library("doParallel")

s1 <- readFASTA(system.file("protseq/P00750.fasta", package = "protr"))[[1]]
s2 <- readFASTA(system.file("protseq/P08218.fasta", package = "protr"))[[1]]
s3 <- readFASTA(system.file("protseq/P10323.fasta", package = "protr"))[[1]]
s4 <- readFASTA(system.file("protseq/P20160.fasta", package = "protr"))[[1]]
s5 <- readFASTA(system.file("protseq/Q9NZP8.fasta", package = "protr"))[[1]]
set.seed(1010)
plist <- as.list(c(s1, s2, s3, s4, s5)[sample(1:5, 100, replace = TRUE)])
psimmat <- parSeqSimDisk(
  plist,
  cores = 2, batches = 10, verbose = TRUE,
  type = "local", submat = "BLOSUM62"
)
} # }