tryggve

Tryggve

Tryggve is the server used for SCALLOP and also SCALLOP-INF at earlier stage of the analysis.

The document covers various aspects such as software and data.

Modules

Instances are exemplified as follows,

module load anaconda2/4.4.0
module load bedtools/2.28.0
module load bgen/20180807
module load perl/5.24.0 annovar/2019oct24
module load bcftools/1.9
module load emacs/26.1
module load gcta/1.91.0beta
module load intel/redist/2019 intel/perflibs/64/2019 gcc/5.4.0 lapack/3.8.0 R/3.5.3-ICC-MKL rstudio/1.1.453
module load libreoffice/6.0.5.2
module load locuszoom/1.4
module load metal/20180828
module load pandoc/2.1
module load parallel/20190122
module load plink2/1.90beta5.4
module load vcftools/0.1.16
module load xpdf/3.04
export threads=1

e.g.,

pandoc tryggve.md -o tryggve.docx
soffice tryggve.docx
xterm -fa arial -fs 12 -bg black -fg white

Parallel computing

METAL.qsub was intended to run batch jobs under Terascale Open-source Resource and QUEue Manager (TORQUE) controls batch jobs and distributed compute nodes, enabling integration with Moab cluster suite and extending Portable Batch System (PBS) with respect to extend scalability, fault tolerance, and functionality. Information on job scheduling on Computerome is availble from https://www.computerome.dk/display/CW/Batch+System,

However, the SCALLOP securitycloud is a 1-node cloud and we resort to parallel instead, e.g,

module load metal/20110325 parallel/20190122
ls METAL/*.run | parallel --dry-run --env HOME -j8 -C' ' 'metal $HOME/INF/{}'

NB METAL add -1 to the filenames.

SOFTWARE ISSUES

Bash functions

There might be problems with R/3.5.3-ICC-MKL so functions are defined in .bashrc which can be invoked at the start of a session, then one invokes R.3.5.3 when necessary.

# This could be very slow
function R.3.5.3()
{
  export R_LIBS=/data/$USER/R:/services/tools/R/3.5.3-ICC-MKL/lib64/R/library
  module load intel/redist/2019 intel/perflibs/64/2019 gcc/5.4.0 lapack/3.8.0 R/3.5.3-ICC-MKL lapack/3.8.0
  source /data/jinhua/parallel-20190222/bin/env_parallel.bash
  alias R='/services/tools/R/3.5.3-ICC-MKL/bin/R -q $@'
}
# An alternative
function R.3.3.1()
{
  module load intel/redist/2019 intel/perflibs/64/2019 gcc/5.4.0 R/3.3.1-ICC-MKL
}

We also use R 3.3.x below.

Code extraction

This is possible with codedown, as noted here.

LocusZoom 1.4

To use LocusZoom 1.4, one only needs to start with

module load gcc/5.4.0
module load R/3.2.5
module load anaconda2/4.4.0
module load locuszoom/1.4

finemap/ldstore

finemap 1.3.1 and ldstore 1.1 are available from /data/jinhua and can be enabled as follows,

ln -fs /data/jinhua/ldstore_v1.1_x86_64/ldstore $HOME/bin/ldstore_v1.1
ln -fs /data/jinhua/finemap_v1.3.1_x86_64/finemap_v1.3.1_x86_64 $HOME/bin/finemap_v1.3.1

finemap 1.4 and ldstore 2.0b are available from /data/jinhua/finemap-1.4.

qctool

TRYGGVE now fixed issue with qctool/2.0.1 for lack of lapack shared libraries as in /data/jinhua/lapack-3.8.0/ and its installation described on GitHub repository, https://github.com/jinghuazhao/Computational-Statistics.

It is then possible to issue module load gcc/5.4.0 lapack/3.8.0 qctool/2.0.1.

NEW SOFTWARE

There are a number of software updates/additions which are worthy of note.

GCTA

A more recenve version is available from /data/jinhua/gcta_1.91.7beta/. This version can handle chi-squared statistics instead of p values in the joint/conditional (COJO) analysis; it also allows for --grm file --pca --out file, i.e., same file root.

GNU parallel

The latest version parallel-20190222 has new features and use them directly on TRYGGVE without invoking modules as follows,

export src=/data/jinhua/parallel-20190222/bin
for i in $(ls $src); do ln -fs $src/$i $HOME/bin/$i; done
export MANPATH=/data/jinhua/parallel-20190222/share/man:$MANPATH

The last line enables man parallel and info parallel.

ImageMagick

This is version 7.0.8-22, made available due to the inability to use the imagemagick/7.0.8-16 module, e.g.,

export PATH=/data/jinhua/ImageMagick-7.0.8-22/bin:$PATH
convert OPG.lz-1.png -resize 130% OPG.lz-3.png
convert \( OPG.qq.png -append OPG.manhattan.png -append OPG.lz-3.png -append \) +append OPG-qml.png

used to generate the figure in the front page. Another very useful utility is its display.

METAL

The version as in /data/jinhua/METAL-2018-08-28 contains modification which allows for CUSTOMVARIABLE to use integer position rather than scientific format as in software-notes.

An extension was made with the Direction column so that if a variant with '+' and '-' effects becomes 'p' and 'n', respectively with P <= 0.05.

To avoid loading the default /usr/bin/metal, one can add

ln -sf /data/jinhua/METAL-2018-08-28/metal $HOME/bin/metal
export PATH=$HOME/bin:/data/jinhua/ImageMagick-7.0.8-22/bin:$PATH

into $HOME/.bashrc.

OpenVPN

In case you see message "All TAP-Windows aapters on this system are currently in use", read this post, esp. install tap-windows-8.21.2.

The log is usually located at "%USERPROFILE%/OpenVPN/log".

R

EasyQC/EasyStrata

The version is 18.1 rather than 9.2 and 8.6 currently online, https://www.uni-regensburg.de/medizin/epidemiologie-praeventivmedizin/genetische-epidemiologie/software/.

gap

First, run R.3.5.3 as defined above.

This version contains functions cis.vs.trans.classification, gc.lambda, invnormal, and here is the way to go

module load intel/redist/2019 intel/perflibs/64/2019 gcc/5.4.0 R/3.5.3-ICC-MKL
tar xvfz gap_1.2.2.tar.gz
cd gap/src
gcc -I/services/tools/intel/perflibs/2019/compilers_and_libraries/linux/mpi/intel64/include -L/services/tools/intel/perflibs/2019/compilers_and_libraries/linux/mpi/intel64/lib/release -L/services/tools/intel/perflibs/2019/compilers_and_libraries/linux/mpi/intel64/lib -Xlinker --enable-new-dtags -Xlinker -rpath -Xlinker /services/tools/intel/perflibs/2019/compilers_and_libraries/linux/mpi/intel64/lib/release -Xlinker -rpath -Xlinker /services/tools/intel/perflibs/2019/compilers_and_libraries/linux/mpi/intel64/lib -Xlinker -rpath -Xlinker /opt/intel/mpi-rt/2017.0.0/intel64/lib/release -Xlinker -rpath -Xlinker /opt/intel/mpi-rt/2017.0.0/intel64/lib -lmpifort -lmpi -ldl -lrt -lpthread -L/services/tools/intel/perflibs/2019//compilers_and_libraries_2019.0.117/linux/mpi/intel64/libfabric/lib -fPIC -c *.c *.f
gcc -shared -L/services/tools/R/3.5.3-ICC-MKL/lib64/R/lib -L/usr/local/lib64 -o gap.so 2k.o 2ld.o cline.o gcontrol_c.o gcx.o gif_c.o hap_c.o hwe.hardy.o kin.morgan.o makeped_c.o mia.o muvar.o package_native_routine_registration_skeleton.o pfc.o pfc.sim.o pgc_c.o whscore.o -L/usr/lib/gcc/x86_64-redhat-linux/4.8.2 -lgfortran -lm -lquadmath -L/services/tools/R/3.5.3-ICC-MKL/lib64/R/lib -lR
cd -
R CMD INSTALL gap -l /data/jinhua/R

The rather imtimidating compiler flags are derived from mpiicc -show as with initial pass requested for libfabric.so.1, and the idea is to get around check for icc which does not exist on the system. To faciliate compiling, gap.sh contains these lines for use.

To take advantage of the circos.mhtplot function, package R/gap.datasets is also made available. The following is an example for cross-check,

R --no-save -q <<END
  gz <- "sumstats/STABILITY/STABILITY.IFN.gamma.gz"
  STABILITY <-  read.table(gz,as.is=TRUE,header=TRUE,sep="\t")
  summary(STABILITY)
  library("GenABEL")
  estlambda(with(STABILITY,PVAL), method="median")
END

Somehow R/gap thus compiled was very slow; so attempt was also made with R 3.3.1 via gap.3.3.1.sh which requires registration of routine family (defined in pfc.f) altered to family_ and the change is made as default.

The fp() function in analysis.sh requires R.3.3.1 to be called.

QCGWAS

The version contains fix to the use of HapMap reference as in software-notes above. The HapMap data as with code from the packages's quick guide is /data/jinhua/data/QCGWAS.

A SUMMARY OF FILES

File specification	Descrription
doc/	Oringal documents
-- KORA.prot.preproc.belowlod.v2.R*	R code for data preprocessing
-- kora.normalised.prot.txt*	sex, age, normalised proteins
-- KORA.pc.below.llod.pdf*	llod check
METAL/	METAL/output scripts by protein
sumstats/	File lists and study directories
tryggve/	Auxiliary files
-- list.sh	Generation of file list and directory
-- format.sh	Code for format GWAS summary statistics
-- lz14.sh	Code to extract LocusZoom 1.4 databases for analysis.sh
-- analysis.sh	Bash code for analysis calling analysis.ini
-- metal.sh	Generation/execution of METAL scripts
-- METAL.qsub	TORQUE qsub script for METAL
-- INTERVAL.sh	INTERVAL analysis
-- KORA.sh	Bash/R scripts to hand KORA data
-- KORA.txt	obsolete version to use simulated data
-- KORA.R	R code to simulate phenotypes
-- EURLD.bed	approximately independent LD blocks for Europeans.
-- QCGWAS.sh	QCGWAS for specific proteins, calling QCGWAS.R
-- snpstats.sh	sumstats/qctool -snp-stats summary, see also qctool.txt
tryggve.md	This document

* from Jimmy

In total, 92 proteins are expected as given in olink.prot.list.txt.

• BioFinder. 91 (no BDNF) proteins. sumstats file named after genes and converted to protein names.
• NSPHS. 91 (no BDNF) proteins, originally a tar.gz file is unpacked into $HOME/INF/work leading to 10 proteins
• EGCUT. 91 (no BDNF) proteins, orginally only 18 proteins though stratified by chromsomes
• INTERVAL. 92 proteins. raw SNPTEST output with information such as info/chip SNPs to be added
• KORA. 91 (no BDNF) proteins, age, sex and individual level imputed genotypes
• LifeLinesDeep. Only 1/25 proteins (unused)
• MadCam. 91 (no IL.6) proteins
• ULSAM. 25 proteins (unused)
• PIVUS. 23 proteins (unused)
• ORCADES. 91 (no BDNF) protein results are available but adding CCL3 which overlaps with MMP.1
• RECOMBINE. 91 (no BDNF) protein results are available with information as described in the analysis plan
• VIS. 91 (no BDNF) protein restults as with ORCADES
• STABILITY. 90 (no BDNF, IL.2) protein.
• STANLEY. 91 (no BDNF) largely complete protein results for lah1 and swe6

NOTEs on results

The directory contains several versions of results,

Directory	Version of results
CEU	Results excluding 18 regions in high LD
HLA	Similar to CEU/ bove but keeps HLA
nold	No exclusion of regions
INTERVAL_nold	nold/ result for INTERVAL

The PLINK --clump-r2 0 results are included here for comparison. Comprehensive information is contained in .cis.vs.trans and cis/trans classificaiton in .out.