Refactored codebase

[jeff-mandell]

Hi, here's a summary of the contents of the pull request:

• I organized scripts and data in accordance with R pacakge conventions. Installation of BISCUT along with all R/Python dependencies is, with luck, 1-2 lines:

	# install remotes package if not already installed
	install.packages('remotes')

	# install the forked version and any dependencies
	remotes::install_github('jeff-mandell/BISCUT-py3')

	# after merging the pull request, installation would be 
	# remotes::install_github('beroukhim-lab/BISCUT-py3')

Please let me know if this doesn't work for you. It's supposed to work, but on one system that I tested, I needed to force re-installation of miniconda in order to get BISCUT's Python dependencies to install correctly (via reticulate::install_miniconda(force = T)).

• I wrapped the BISCUT workflow into R functions so that the user can do everything from the R environment. The worflow is to call make_breakpoint_files() to process a segment file (such as PANCAN_ISAR.seg.txt) and produce a breakpoint file directory, and then to call do_biscut() with the
  breakpoint file directory as an input. Both functions can use parallelization when the cores argument is specified. I made some other optimizations to speed runtime. Besides printing output files, I also made the main function return peak information within R. Here is example usage:

# Assumes segment file exists in working directory.
library(BISCUT)
make_breakpoint_files(segment_file = 'PANCAN_ISAR.seg.txt', output_dir = 'my_breakpoints', cores = 8)
results = do_biscut(breakpoint_file_dir = 'my_breakpoints', results_dir = 'my_biscut_output', cores = 8)

# see ?do_biscut for documentation of output and function arguments

• The output directory and reference files (abslocs, genelocs) can now be user-specified, as can parameters that could previously only be changed by editing source code (output directory, telcent_thres, amplitude_threshold, confidence interval, n bootstraps). The previously hardcoded data and parameter values are used as defaults when unspecified by the user.

• The code was doing a lot of error suppression which could prevent the user from knowing about potential problems. I removed most of this and resolved any errors and warnings. There are probably still some that could be triggered by situations or system configurations that I didn't test.

Questions/issues

• There appeared to be an error in the chromosome coordinates file (now at inst/extdata/SNP6_hg19_chromosome_locs_200605.txt). Every chromosome has q_start1 = centromere + offset except for chr15, where instead q_start1 = q_end1 (= 2409817111). It appears that the q_start1 value was accidentally copied from q_end1, unless there is something I don't understand about the format. I changed the value to match the other chromosomes. Please let me know if it should be changed back.

• FYI, process_for_ggplot_jagged() doesn't run, partly because it wants to load files ending in plotpeaks.txt. I couldn't find any code in the project that generates such files. Therefore, do_biscut() currently just calls process_for_ggplot().

• plot_fig_2() is runnable, but the plots look odd. I couldn't tell which plot(s) from the paper they are supposed to resemble. If there isn't a continuing need for the function, I could remove the filter_BISCUT_arms() call from the main function to avoid generating unused files.

• Previously, there were hard-coded genes of interest--c('CDKN2A','TERT','MYC','BAP1','TERC','TP53','ARID1A','EGFR','PPM1D')--in BISCUT_peak_finding.R. I removed this declaration, as the user can instead call filter_BISCUT_knowngenes (via an R wrapper function, extract_gene_results()) on any desired genes. However, since process_for_ggplot_jagged() isn't running, I am not sure of the use case for filter_BISCUT_knowngenes().

Suggestions:

• Fill in the placeholder information in the package's top-level DESCRIPTION file: title, description, and contributors. The conventions for filling in contributors are described here: https://r-pkgs.org/description.html#sec-description-authors-at-r.
• Also in the DESCRIPTION, I temporarily put in version 1.2.0.9000 to indicate a development version. This should be changed to 1.3.0 (or 2.0.0, if you are into semantic versioning, or any other version number you like).
• The README should be updated. I copied a lot of the descriptions of files and parameters into the R scripts, so that the information appears when a user calls ?do_biscut() or ?make_breakpoint_files(). Therefore, some of this information could be removed from the README. If you think it is a good idea, a compressed segment file, perhaps covering fewer samples than the PANCAN file, could be saved in the repository so that the README could give runnable code on an example segment file.
• Most of the data from do_biscut() is currently both returned within R and printed to output files, in somewhat different formats. If you are interested, I could further prune down output data and make all output file writing optional. I could also reduce the breakpoint file directory to a single breakpoint file, and make do_biscut() accept either the single file or a data.table/data.frame version of it.