chris1221 / corge

Evaluation of Simultaneous Inference Methods for the Human Genome.

Hi there, I'm Chris.

I am currently a lead scientist at Deep Genomics in Toronto. My work involves using machine learning and AI to identify causal drug targets from human genetics data. I previously completed my DPhil working with Gerton Lunter and Thomas Milne at the University of Oxford. Check out my website for a breakdown of previous projects and repositories of interest.

corge's Issues

Remedy .GlobalEnv() workaround for `samp`

Vectorize SNPTEST2R.R

using foreach

Import shell scripts from inst/bash in sim.gen()

Adjust p values by grouping vector [correct() and analyze()]

Need to alter correct.R or analyze.new.R to account for the fact that grouping variables may be included.

Issues Remaining

January 12 Meeting

• The clusters aren't obviously separable, so maybe add as a third dimension a density peak to see if the groups really are different in the following:

• Incorporate true positive gradient instead of absolute number of TPs
• Sweep over concentrations for cluster enrichment
• UK10K Data access

Clean up namespace conflict warnings with @ImportFrom

data.table::fread() / coR-ge::gen2r() => readlines() + foreach::dopar()

Unsigned to signed int comparison [-wSign-compare]

ld.cpp:38:21: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
  for(uword i = 0; i < nCausal; i++){

Probably okay. This comes from the counter in RcppArmadillo.

base::for => doMC::foreach [Oxford to R format conversion]

Convert gen2r.R to use better looping constructs. Additionally check to see number of threads avvailble to multicore.

Failed attempts at speeding up reading (i.e. things which are not faster than fread).

Attempts at speeding up the reading and converting of .gen files.

This thread exists as a warning and reminder to myself of how truly awful I am at programming.

Sub Script function

• Create cluster submission script within inst/bash
• Create wrapped function for sub so don't have to leave R, ie

if(!require(pacman)) install.packages("pacman", repos = uT)
p_load(foreach, SnowMC)

install_github("Chris1221", "coRge")

uT = cran repo
foreach(i = 1:10) %:% foreach(j = 1:10) %do% sub(i,j) #from within interactive session

snps out of scope in coRge::phen

[v0.6.4.9000] Deprecate i and j args in favour of path to file / dir

Logging with futile.logger and pass up threshold

Definitely need to do this, but maybe not by Monday. Might be as simple as replacing message( with flog.info( but check this first.

Non finite elements introduced in effect size calculation

CentOS 6 Compatibility (g++ < 4.6)

Oh boy. This is going to be a headache.

Export group and group_name arguements

Vignette

Add in gene "mode"

Mode arguement must be changed in analyze.R along with correct.R and stratify.R.

Another option would be to harmonize the entire function, but I don't think I have the time right now.

[v0.6.3.9000] Singular Matrices in c(1, gen)

Dealing with this by Moore-Penrose pseudo-inverse of matrix.

arma::pinv()

Circumvent need for coRge::gen2r()

This is way too long, maybe create a vector with three entries for each es and simply multiply the matrices. This has to be faster. Have to make sure that 0*x = 0 and that the 0s don't contribute to rowsums. they shouldn't but just make sure.