Fortran can serve as a valuable tool for heavy numerical calculations
in Next Generation Sequencing (NGS) analysis, which may often involve
reading and writing many large .bed files in succession. The
IO-Fortran-Library is optimized for both performance and memory
consumption when reading and writing large text files, streamlining
performance for Fortran bioinformatics applications.
To demonstrate the speed of file I/O, the following program reads the
Twist Human Core Exome target .bed file for hg38 obtained from
Twist Bioscience
into a cell array and then writes the cell array to a new file in a
round-trip, comparing the two files for an exact match and providing
the total time elapsed:
program main
use, intrinsic :: iso_fortran_env, only: int64, real64, compiler_version
use io_fortran_lib, only: String, str, TAB, operator(+), operator(==)
implicit none (type, external)
type(String) :: hg38, hg38_new
type(String), allocatable :: cells(:,:)
integer(int64) :: t1, t2
real(real64) :: wall_time, rate
call system_clock(t1)
call hg38%read_file("./data/hg38.bed", cell_array=cells, column_separator=TAB)
call hg38_new%write_file(cells, "./data/hg38_new.bed", column_separator=TAB)
call system_clock(t2, count_rate=rate); wall_time = real(t2-t1,real64)/rate
write(*,"(a,l)") "New file and original are exact match: ", hg38_new == hg38
write(*,"(a)") "Wall time: " + str(wall_time, fmt="f", decimals=3) + " s " + &
'using compiler: "' + compiler_version() + '".'
end program main
The file hg38.bed is provided locally in /data and contains
192262 lines of TAB-delimited data.
Note
With the Intel Fortran compiler ifx, we may need to specify
-heap-arrays 0 to avoid a segmentation fault when reading a file of
this size, as noted in compiler-dependent
behavior.