quality-control.make

# FASTQ QC# {{{
.PHONY: fastq-qc
fastq-qc: ${fastqc_zip}

${data_path}/fastqc/%_fastqc.zip: ${data_path}/%.fastq.gz
	${bsub} -n 6 -R 'rusage[mem=6000]' "run-fastqc.sh $<"

# }}}

# ChIP quality control# {{{
.PHONY: chip-qc
## Calculates cross-correlation and makes custom fingerprint plots to investigate enrichment
chip-qc: ${cross_correlation} ${enrichment_files} ${enrichment_plots}

# use 6 nodes, savp creates the cross correlation plot
# run_spp_parallel.R is a modified version of phantompeaktools/run_spp.R to use 
# makeForkCluster from parallel package (instead of makeCluster from snow, which doesn't play well
# with LSF in this case). Similar problem as here: https://stat.ethz.ch/pipermail/r-sig-hpc/2012-December/001568.html

${qc_path}%.cross-correlation:: ${map_path}%.bam
	${bsub} -n 8 -M ${memlimit} -R "select[ncores=8]" "Rscript ~/local/phantompeakqualtools/run_spp_parallel.R -p=8 -c=$< -savp -tmpdir=${qc_path}/tmp -out=$@"
	mv ${map_path}$*.pdf ${qc_path}

${qc_path}/%.cross-correlation:: ${map_path}/ddup/%.bam
	${bsub} -n 8 -M ${memlimit} -R "select[ncores=8]" "Rscript ~/local/phantompeakqualtools/run_spp_nodups.R -p=8 -c=$< -savp -tmpdir=${qc_path}/tmp -out=$@"
	mv ${map_path}/ddup/$*.pdf ${qc_path}

${qc_path}%.enrichment: ${map_path}%.bam
	${bsub} -n 4 "perl ~/source/check-chip-enrichment.pl ${reference}/${assembly}.fa.fai 1000 $< > $@"

${qc_path}/plots/%.png: ${qc_path}/%.enrichment
	$(eval target_ctrl := $(call KEEP,$(call get-base,$*),$(call KEEP,$(call get-ctrl-type,$*)_,${enrichment_files})))
	${bsub} -M ${memlimit} "Rscript ~/source/plot-chip-enrichment.R -b $< $(target_ctrl) -o $(dir $(@D))"

# }}}

# Correlation # {{{
# ${sample_correlation} needs to be provided from the project specific makefile, 
# where any particular filtering is done.
correlation_sets := $(patsubst %.pdf, %, ${sample_correlation})

define COR

$(1).sample-correlation: $(call KEEP,ddup,$(call KEEP,$(notdir $(1)),${bigwig}))
#	@echo $$^ | tr ' ' '\n' > $(1).tmp
	@echo $$^ | tr ' ' '\n' #> $(1).tmp
#	Rscript -e "a=read.delim('$(1).tmp', stringsAsFactors=F); a=a[[1]]; b=as.data.frame(t(combn(a,2))); write.table(b, '$(1).pairwise-combn', col.names=F, row.names=F, sep=' ', quote=F)"
#	${bsub} run-bigWigCorrelate-on-pairs.sh $(1).pairwise-combn $(1).sample-correlation
#	rm $(1).tmp

$(1).pdf: $(1).sample-correlation
	${bsub} "Rscript ~/source/plot-bw-correlation.R -s $(1).sample-correlation"

endef

$(foreach cn,${correlation_sets},$(eval $(call COR,$(cn))))

.PHONY: correlation
## Calculate correlation of samples genome wide (not on specific regions)
correlation: ${sample_correlation} #${correlation_plots}
# }}}
# vim: ft=make