Last updated: 2019-11-20

Checks: 7 0

Knit directory: Comparative_APA/analysis/

This reproducible R Markdown analysis was created with workflowr (version 1.5.0). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.


Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.

Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.

The command set.seed(20190902) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.

Great job! Recording the operating system, R version, and package versions is critical for reproducibility.

Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.

Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.

Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility. The version displayed above was the version of the Git repository at the time these results were generated.

Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated:


Ignored files:
    Ignored:    .DS_Store
    Ignored:    .Rhistory
    Ignored:    .Rproj.user/
    Ignored:    code/chimp_log/
    Ignored:    code/human_log/
    Ignored:    data/.DS_Store
    Ignored:    data/RNASEQ_metadata.txt.sb-51f67ae1-HXp7Gq/
    Ignored:    data/RNASEQ_metadata_2Removed.txt.sb-4426323c-a4lBwx/
    Ignored:    data/metadata_HCpanel.txt.sb-a3d92a2d-b9cYoF/
    Ignored:    data/metadata_HCpanel.txt.sb-f4823d1e-qihGek/

Untracked files:
    Untracked:  ._.DS_Store
    Untracked:  Chimp/
    Untracked:  Human/
    Untracked:  analysis/assessReadQual.Rmd
    Untracked:  code/._Config_chimp.yaml
    Untracked:  code/._Config_human.yaml
    Untracked:  code/._DiffSplice.sh
    Untracked:  code/._DiffSplicePlots.sh
    Untracked:  code/._DiffSplicePlots_gencode.sh
    Untracked:  code/._DiffSplice_gencode.sh
    Untracked:  code/._LiftOrthoPAS2chimp.sh
    Untracked:  code/._MapBadSamples.sh
    Untracked:  code/._Snakefile
    Untracked:  code/._SnakefilePAS
    Untracked:  code/._SnakefilePASfilt
    Untracked:  code/._SortIndexBadSamples.sh
    Untracked:  code/._bed215upbed.py
    Untracked:  code/._bed2SAF_gen.py
    Untracked:  code/._buildIndecpantro5
    Untracked:  code/._buildIndecpantro5.sh
    Untracked:  code/._buildStarIndex.sh
    Untracked:  code/._cleanbed2saf.py
    Untracked:  code/._cluster.json
    Untracked:  code/._cluster2bed.py
    Untracked:  code/._clusterLiftReverse.sh
    Untracked:  code/._clusterLiftprimary.sh
    Untracked:  code/._converBam2Junc.sh
    Untracked:  code/._extraSnakefiltpas
    Untracked:  code/._filter5percPAS.py
    Untracked:  code/._filterPASforMP.py
    Untracked:  code/._filterPostLift.py
    Untracked:  code/._fixExonFC.py
    Untracked:  code/._fixUTRexonanno.py
    Untracked:  code/._formathg38Anno.py
    Untracked:  code/._formatpantro6Anno.py
    Untracked:  code/._intersectLiftedPAS.sh
    Untracked:  code/._liftPAS19to38.sh
    Untracked:  code/._makeSamplyGroupsHuman_TvN.py
    Untracked:  code/._mapRNAseqhg19.sh
    Untracked:  code/._maphg19.sh
    Untracked:  code/._maphg19_subjunc.sh
    Untracked:  code/._mergedBam2BW.sh
    Untracked:  code/._nameClusters.py
    Untracked:  code/._overlapapaQTLPAS.sh
    Untracked:  code/._prepareCleanLiftedFC_5perc4LC.py
    Untracked:  code/._preparePAS4lift.py
    Untracked:  code/._primaryLift.sh
    Untracked:  code/._processhg38exons.py
    Untracked:  code/._quantJunc.sh
    Untracked:  code/._recLiftchim2human.sh
    Untracked:  code/._revLiftPAShg38to19.sh
    Untracked:  code/._reverseLift.sh
    Untracked:  code/._runChimpDiffIso.sh
    Untracked:  code/._runHumanDiffIso.sh
    Untracked:  code/._runNuclearDifffIso.sh
    Untracked:  code/._run_chimpverifybam.sh
    Untracked:  code/._run_verifyBam.sh
    Untracked:  code/._snakemake.batch
    Untracked:  code/._snakemakePAS.batch
    Untracked:  code/._snakemakePASchimp.batch
    Untracked:  code/._snakemakePAShuman.batch
    Untracked:  code/._snakemake_chimp.batch
    Untracked:  code/._snakemake_human.batch
    Untracked:  code/._snakemakefiltPAS.batch
    Untracked:  code/._snakemakefiltPAS_chimp
    Untracked:  code/._snakemakefiltPAS_chimp.sh
    Untracked:  code/._snakemakefiltPAS_human.sh
    Untracked:  code/._submit-snakemake-chimp.sh
    Untracked:  code/._submit-snakemake-human.sh
    Untracked:  code/._submit-snakemakePAS-chimp.sh
    Untracked:  code/._submit-snakemakePAS-human.sh
    Untracked:  code/._submit-snakemakefiltPAS-chimp.sh
    Untracked:  code/._submit-snakemakefiltPAS-human.sh
    Untracked:  code/._subset_diffisopheno_Nuclear_HvC.py
    Untracked:  code/._transcriptDTplotsNuclear.sh
    Untracked:  code/._verifyBam4973.sh
    Untracked:  code/._verifyBam4973inHuman.sh
    Untracked:  code/._wrap_chimpverifybam.sh
    Untracked:  code/._wrap_verifyBam.sh
    Untracked:  code/.snakemake/
    Untracked:  code/Config_chimp.yaml
    Untracked:  code/Config_human.yaml
    Untracked:  code/DiffSplice.err
    Untracked:  code/DiffSplice.out
    Untracked:  code/DiffSplice.sh
    Untracked:  code/DiffSplicePlots.err
    Untracked:  code/DiffSplicePlots.out
    Untracked:  code/DiffSplicePlots.sh
    Untracked:  code/DiffSplicePlots_gencode.sh
    Untracked:  code/DiffSplice_gencode.sh
    Untracked:  code/GencodeDiffSplice.err
    Untracked:  code/GencodeDiffSplice.out
    Untracked:  code/LiftClustersFirst.err
    Untracked:  code/LiftClustersFirst.out
    Untracked:  code/LiftClustersSecond.err
    Untracked:  code/LiftClustersSecond.out
    Untracked:  code/LiftOrthoPAS2chimp.sh
    Untracked:  code/LiftorthoPAS.err
    Untracked:  code/LiftorthoPASt.out
    Untracked:  code/Log.out
    Untracked:  code/MapBadSamples.err
    Untracked:  code/MapBadSamples.out
    Untracked:  code/MapBadSamples.sh
    Untracked:  code/Rev_liftoverPAShg19to38.err
    Untracked:  code/Rev_liftoverPAShg19to38.out
    Untracked:  code/SAF215upbed_gen.py
    Untracked:  code/Snakefile
    Untracked:  code/SnakefilePAS
    Untracked:  code/SnakefilePASfilt
    Untracked:  code/SortIndexBadSamples.err
    Untracked:  code/SortIndexBadSamples.out
    Untracked:  code/SortIndexBadSamples.sh
    Untracked:  code/TotalTranscriptDTplot.err
    Untracked:  code/TotalTranscriptDTplot.out
    Untracked:  code/Upstream10Bases_general.py
    Untracked:  code/apaQTLsnake.err
    Untracked:  code/apaQTLsnake.out
    Untracked:  code/apaQTLsnakePAS.err
    Untracked:  code/apaQTLsnakePAS.out
    Untracked:  code/apaQTLsnakePAShuman.err
    Untracked:  code/bam2junc.err
    Untracked:  code/bam2junc.out
    Untracked:  code/bed215upbed.py
    Untracked:  code/bed2SAF_gen.py
    Untracked:  code/bed2saf.py
    Untracked:  code/bg_to_cov.py
    Untracked:  code/buildIndecpantro5
    Untracked:  code/buildIndecpantro5.sh
    Untracked:  code/buildStarIndex.sh
    Untracked:  code/callPeaksYL.py
    Untracked:  code/chooseAnno2Bed.py
    Untracked:  code/chooseAnno2SAF.py
    Untracked:  code/cleanbed2saf.py
    Untracked:  code/cluster.json
    Untracked:  code/cluster2bed.py
    Untracked:  code/clusterLiftReverse.sh
    Untracked:  code/clusterLiftprimary.sh
    Untracked:  code/clusterPAS.json
    Untracked:  code/clusterfiltPAS.json
    Untracked:  code/converBam2Junc.sh
    Untracked:  code/convertNumeric.py
    Untracked:  code/environment.yaml
    Untracked:  code/extraSnakefiltpas
    Untracked:  code/filter5perc.R
    Untracked:  code/filter5percPAS.py
    Untracked:  code/filter5percPheno.py
    Untracked:  code/filterBamforMP.pysam2_gen.py
    Untracked:  code/filterMissprimingInNuc10_gen.py
    Untracked:  code/filterPASforMP.py
    Untracked:  code/filterPostLift.py
    Untracked:  code/filterSAFforMP_gen.py
    Untracked:  code/filterSortBedbyCleanedBed_gen.R
    Untracked:  code/filterpeaks.py
    Untracked:  code/fixExonFC.py
    Untracked:  code/fixFChead.py
    Untracked:  code/fixFChead_bothfrac.py
    Untracked:  code/fixUTRexonanno.py
    Untracked:  code/formathg38Anno.py
    Untracked:  code/generateStarIndex.err
    Untracked:  code/generateStarIndex.out
    Untracked:  code/generateStarIndexHuman.err
    Untracked:  code/generateStarIndexHuman.out
    Untracked:  code/intersectAnno.err
    Untracked:  code/intersectAnno.out
    Untracked:  code/intersectLiftedPAS.sh
    Untracked:  code/liftPAS19to38.sh
    Untracked:  code/liftoverPAShg19to38.err
    Untracked:  code/liftoverPAShg19to38.out
    Untracked:  code/log/
    Untracked:  code/make5percPeakbed.py
    Untracked:  code/makeFileID.py
    Untracked:  code/makePheno.py
    Untracked:  code/makeSamplyGroupsChimp_TvN.py
    Untracked:  code/makeSamplyGroupsHuman_TvN.py
    Untracked:  code/mapRNAseqhg19.sh
    Untracked:  code/maphg19.err
    Untracked:  code/maphg19.out
    Untracked:  code/maphg19.sh
    Untracked:  code/maphg19_sub.err
    Untracked:  code/maphg19_sub.out
    Untracked:  code/maphg19_subjunc.sh
    Untracked:  code/mergedBam2BW.sh
    Untracked:  code/mergedbam2bw.err
    Untracked:  code/mergedbam2bw.out
    Untracked:  code/nameClusters.py
    Untracked:  code/namePeaks.py
    Untracked:  code/nuclearTranscriptDTplot.err
    Untracked:  code/nuclearTranscriptDTplot.out
    Untracked:  code/overlapPAS.err
    Untracked:  code/overlapPAS.out
    Untracked:  code/overlapapaQTLPAS.sh
    Untracked:  code/peak2PAS.py
    Untracked:  code/pheno2countonly.R
    Untracked:  code/prepareCleanLiftedFC_5perc4LC.py
    Untracked:  code/preparePAS4lift.py
    Untracked:  code/prepare_phenotype_table.py
    Untracked:  code/primaryLift.err
    Untracked:  code/primaryLift.out
    Untracked:  code/primaryLift.sh
    Untracked:  code/processhg38exons.py
    Untracked:  code/quantJunc.sh
    Untracked:  code/quantLiftedPAS.err
    Untracked:  code/quantLiftedPAS.out
    Untracked:  code/quantLiftedPAS.sh
    Untracked:  code/quatJunc.err
    Untracked:  code/quatJunc.out
    Untracked:  code/recChimpback2Human.err
    Untracked:  code/recChimpback2Human.out
    Untracked:  code/recLiftchim2human.sh
    Untracked:  code/revLift.err
    Untracked:  code/revLift.out
    Untracked:  code/revLiftPAShg38to19.sh
    Untracked:  code/reverseLift.sh
    Untracked:  code/runChimpDiffIso.sh
    Untracked:  code/runHumanDiffIso.sh
    Untracked:  code/runNuclearDifffIso.sh
    Untracked:  code/run_Chimpleafcutter_ds.err
    Untracked:  code/run_Chimpleafcutter_ds.out
    Untracked:  code/run_Chimpverifybam.err
    Untracked:  code/run_Chimpverifybam.out
    Untracked:  code/run_Humanleafcutter_ds.err
    Untracked:  code/run_Humanleafcutter_ds.out
    Untracked:  code/run_Nuclearleafcutter_ds.err
    Untracked:  code/run_Nuclearleafcutter_ds.out
    Untracked:  code/run_chimpverifybam.sh
    Untracked:  code/run_verifyBam.sh
    Untracked:  code/run_verifybam.err
    Untracked:  code/run_verifybam.out
    Untracked:  code/slurm-62824013.out
    Untracked:  code/slurm-62825841.out
    Untracked:  code/slurm-62826116.out
    Untracked:  code/snakePASChimp.err
    Untracked:  code/snakePASChimp.out
    Untracked:  code/snakePAShuman.out
    Untracked:  code/snakemake.batch
    Untracked:  code/snakemakeChimp.err
    Untracked:  code/snakemakeChimp.out
    Untracked:  code/snakemakeHuman.err
    Untracked:  code/snakemakeHuman.out
    Untracked:  code/snakemakePAS.batch
    Untracked:  code/snakemakePASFiltChimp.err
    Untracked:  code/snakemakePASFiltChimp.out
    Untracked:  code/snakemakePASFiltHuman.err
    Untracked:  code/snakemakePASFiltHuman.out
    Untracked:  code/snakemakePASchimp.batch
    Untracked:  code/snakemakePAShuman.batch
    Untracked:  code/snakemake_chimp.batch
    Untracked:  code/snakemake_human.batch
    Untracked:  code/snakemakefiltPAS.batch
    Untracked:  code/snakemakefiltPAS_chimp.sh
    Untracked:  code/snakemakefiltPAS_human.sh
    Untracked:  code/submit-snakemake-chimp.sh
    Untracked:  code/submit-snakemake-human.sh
    Untracked:  code/submit-snakemakePAS-chimp.sh
    Untracked:  code/submit-snakemakePAS-human.sh
    Untracked:  code/submit-snakemakefiltPAS-chimp.sh
    Untracked:  code/submit-snakemakefiltPAS-human.sh
    Untracked:  code/subset_diffisopheno.py
    Untracked:  code/subset_diffisopheno_Chimp_tvN.py
    Untracked:  code/subset_diffisopheno_Huma_tvN.py
    Untracked:  code/subset_diffisopheno_Nuclear_HvC.py
    Untracked:  code/transcriptDTplotsNuclear.sh
    Untracked:  code/transcriptDTplotsTotal.sh
    Untracked:  code/verifyBam4973.sh
    Untracked:  code/verifyBam4973inHuman.sh
    Untracked:  code/verifybam4973.err
    Untracked:  code/verifybam4973.out
    Untracked:  code/verifybam4973HumanMap.err
    Untracked:  code/verifybam4973HumanMap.out
    Untracked:  code/wrap_Chimpverifybam.err
    Untracked:  code/wrap_Chimpverifybam.out
    Untracked:  code/wrap_chimpverifybam.sh
    Untracked:  code/wrap_verifyBam.sh
    Untracked:  code/wrap_verifybam.err
    Untracked:  code/wrap_verifybam.out
    Untracked:  data/._.DS_Store
    Untracked:  data/._RNASEQ_metadata.txt
    Untracked:  data/._RNASEQ_metadata.txt.sb-51f67ae1-HXp7Gq
    Untracked:  data/._RNASEQ_metadata.xlsx
    Untracked:  data/._RNASEQ_metadata_2Removed.txt
    Untracked:  data/._RNASEQ_metadata_2Removed.txt.sb-4426323c-a4lBwx
    Untracked:  data/._RNASEQ_metadata_2Removed.xlsx
    Untracked:  data/._metadata_HCpanel.txt
    Untracked:  data/._metadata_HCpanel.txt.sb-a3d92a2d-b9cYoF
    Untracked:  data/._metadata_HCpanel.txt.sb-f4823d1e-qihGek
    Untracked:  data/._metadata_HCpanel.xlsx
    Untracked:  data/._~$RNASEQ_metadata.xlsx
    Untracked:  data/._~$metadata_HCpanel.xlsx
    Untracked:  data/CompapaQTLpas/
    Untracked:  data/DTmatrix/
    Untracked:  data/DiffIso_Nuclear/
    Untracked:  data/DiffSplice/
    Untracked:  data/MapStats/
    Untracked:  data/NuclearHvC/
    Untracked:  data/Peaks_5perc/
    Untracked:  data/Pheno_5perc/
    Untracked:  data/Pheno_5perc_nuclear/
    Untracked:  data/Pheno_5perc_total/
    Untracked:  data/RNASEQ_metadata.txt
    Untracked:  data/RNASEQ_metadata.xlsx
    Untracked:  data/RNASEQ_metadata_2Removed.txt
    Untracked:  data/RNASEQ_metadata_2Removed.xlsx
    Untracked:  data/TwoBadSampleAnalysis/
    Untracked:  data/chainFiles/
    Untracked:  data/cleanPeaks_anno/
    Untracked:  data/cleanPeaks_byspecies/
    Untracked:  data/cleanPeaks_lifted/
    Untracked:  data/liftover_files/
    Untracked:  data/metadata_HCpanel.txt
    Untracked:  data/metadata_HCpanel.xlsx
    Untracked:  data/primaryLift/
    Untracked:  data/reverseLift/
    Untracked:  data/~$RNASEQ_metadata.xlsx
    Untracked:  data/~$metadata_HCpanel.xlsx
    Untracked:  output/dtPlots/
    Untracked:  projectNotes.Rmd

Unstaged changes:
    Modified:   analysis/CorrbetweenInd.Rmd
    Modified:   analysis/InvestigateBadSamples.Rmd
    Modified:   analysis/MapRNAhg19.Rmd
    Modified:   analysis/PASnumperSpecies.Rmd
    Modified:   analysis/annotationInfo.Rmd
    Modified:   analysis/diffSplicing.Rmd
    Modified:   analysis/verifyBAM.Rmd

Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.


These are the previous versions of the R Markdown and HTML files. If you’ve configured a remote Git repository (see ?wflow_git_remote), click on the hyperlinks in the table below to view them.

File Version Author Date Message
Rmd db0484c brimittleman 2019-11-21 add PC corr
html 712106e brimittleman 2019-11-19 Build site.
Rmd 8dc9ea0 brimittleman 2019-11-19 first pipeline for de
html 586c9ec brimittleman 2019-11-13 Build site.
Rmd bedfa41 brimittleman 2019-11-13 question PCA methods
html a22bae9 brimittleman 2019-11-13 Build site.
Rmd a52c26d brimittleman 2019-11-13 look at pca and tech factors
html da4bab0 brimittleman 2019-11-12 Build site.
Rmd 98d7f9b brimittleman 2019-11-12 add cpm pca
html 32b435b brimittleman 2019-11-12 Build site.
Rmd 1ce8433 brimittleman 2019-11-12 start normalization
html 2c02d70 brimittleman 2019-11-12 Build site.
Rmd 53642f7 brimittleman 2019-11-12 add mapp stats
html dc91b0a brimittleman 2019-11-11 Build site.
Rmd b5ba82e brimittleman 2019-11-11 add diff expression and diff splicing

library(workflowr)
This is workflowr version 1.5.0
Run ?workflowr for help getting started
library(tidyverse)
── Attaching packages ────────────────────────────────────────────────────────────── tidyverse 1.2.1 ──
✔ ggplot2 3.1.1       ✔ purrr   0.3.2  
✔ tibble  2.1.1       ✔ dplyr   0.8.0.1
✔ tidyr   0.8.3       ✔ stringr 1.3.1  
✔ readr   1.3.1       ✔ forcats 0.3.0  
── Conflicts ───────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
library("scales")

Attaching package: 'scales'
The following object is masked from 'package:purrr':

    discard
The following object is masked from 'package:readr':

    col_factor
library("gplots")

Attaching package: 'gplots'
The following object is masked from 'package:stats':

    lowess
library("edgeR")
Loading required package: limma
library("R.utils")
Loading required package: R.oo
Loading required package: R.methodsS3
R.methodsS3 v1.7.1 (2016-02-15) successfully loaded. See ?R.methodsS3 for help.
R.oo v1.22.0 (2018-04-21) successfully loaded. See ?R.oo for help.

Attaching package: 'R.oo'
The following objects are masked from 'package:methods':

    getClasses, getMethods
The following objects are masked from 'package:base':

    attach, detach, gc, load, save
R.utils v2.7.0 successfully loaded. See ?R.utils for help.

Attaching package: 'R.utils'
The following object is masked from 'package:tidyr':

    extract
The following object is masked from 'package:utils':

    timestamp
The following objects are masked from 'package:base':

    cat, commandArgs, getOption, inherits, isOpen, parse, warnings
library("limma")
library("VennDiagram")
Loading required package: grid
Loading required package: futile.logger
library("RColorBrewer")
library(reshape2)

Attaching package: 'reshape2'
The following object is masked from 'package:tidyr':

    smiths

For this analysis I do preprocessing with the Snakemake pipeline. The snakemake will map the RNA seq and quantify orthologous exons.

From FastQC:

  • Does not look like there is adapter contamination

  • No reads tagged as bad quality

Assess mapping:

metaData=read.table("../data/RNASEQ_metadata.txt", header = T, stringsAsFactors = F)
metaData$Species=as.factor(metaData$Species)
metaData$Collection=as.factor(metaData$Collection)
readInfo=metaData %>% mutate(AAUnMapped= Reads-Mapped, ABNotOrtho= Mapped-AssignedOrtho) %>% select(Line, Species, AAUnMapped, ABNotOrtho, AssignedOrtho) %>%  gather(key="Category", value="Number", -Line, -Species)


ggplot(readInfo, aes(x=Line,y=Number, fill=Category)) + geom_bar(stat="identity") + scale_fill_brewer(palette = "Dark2",name = "Type", labels = c("Unmapped", "Mapped not ortho", "Assigned Ortho Exon"))+theme(axis.text.x = element_text( hjust = 0,vjust = 1, size = 6, angle = 90)) + labs(y="Reads", title="Human and chimp read statistics") 

Version Author Date
32b435b brimittleman 2019-11-12
2c02d70 brimittleman 2019-11-12

Proportion of reads.

readProp=metaData %>% mutate(Aunmapped=1-percentMapped, MappednotOrtho=percentMapped-percentOrtho) %>% select(Line,Species, percentOrtho, MappednotOrtho, Aunmapped) %>%  gather(key="Category", value="Proportion", -Line, -Species)

ggplot(readProp, aes(x=Line,y=Proportion, fill=Category)) + geom_bar(stat="identity") + scale_fill_brewer(palette = "Dark2", name="", labels = c("Unmapped", "Mapped not ortho", "Assigned Ortho Exon"))+theme(axis.text.x = element_text( hjust = 0,vjust = 1, size = 6, angle = 90)) + labs(y="Reads", title="Human and chimp read proportions") 

Version Author Date
32b435b brimittleman 2019-11-12
2c02d70 brimittleman 2019-11-12

By species:

ggplot(readInfo,aes(x=Category, y=Number, by=Species, fill=Species)) + geom_boxplot() +scale_x_discrete( breaks=c("AAUnMapped","ABNotOrtho","AssignedOrtho"),labels=c("Unmapped", "Not in OrthoExon", "Assigned to OrthoExon")) + scale_fill_brewer(palette = "Dark2") + labs(title="Mapped reads by Species", y="Reads", x="")

Version Author Date
2c02d70 brimittleman 2019-11-12
ggplot(readProp,aes(x=Category, y=Proportion, by=Species, fill=Species)) + geom_boxplot() + scale_fill_brewer(palette = "Dark2") + labs(title="Map Proportion by Species", y="Proportion", x="") + scale_x_discrete( breaks=c("Aunmapped","MappedNotOrtho","percentOrtho"),labels=c("Unmapped", "Not in OrthoExon", "Assigned to OrthoExon"))

Version Author Date
2c02d70 brimittleman 2019-11-12

Diffferential Expression

Code originally from Lauren Blake (http://lauren-blake.github.io/Reg_Evo_Primates/analysis/Normalization_plots.html)

Raw Counts

Fix header for fc files:

python fixExonFC.py /project2/gilad/briana/Comparative_APA/Human/data/RNAseq/ExonCounts/RNAseqOrthoExon.fc /project2/gilad/briana/Comparative_APA/Human/data/RNAseq/ExonCounts/RNAseqOrthoExon.fixed.fc

python fixExonFC.py /project2/gilad/briana/Comparative_APA/Chimp/data/RNAseq/ExonCounts/RNAseqOrthoExon.fc /project2/gilad/briana/Comparative_APA/Chimp/data/RNAseq/ExonCounts/RNAseqOrthoExon.fixed.fc
HumanCounts=read.table("../Human/data/RNAseq/ExonCounts/RNAseqOrthoExon.fixed.fc", header = T, stringsAsFactors = F) %>% select(-Chr,-Start,-End,-Strand, -Length)

ChimpCounts=read.table("../Chimp/data/RNAseq/ExonCounts/RNAseqOrthoExon.fixed.fc", header = T, stringsAsFactors = F) %>% select(-Chr,-Start,-End,-Strand, -Length)


counts_genes=HumanCounts %>% inner_join(ChimpCounts,by="Geneid") %>% column_to_rownames(var="Geneid")

head(counts_genes)
                NA18504 NA18510 NA18523 NA18498 NA18499 NA18502 NAPT30
ENSG00000188976      24      50      31      69      34      58      2
ENSG00000188157     106      65     106      12     128      39      5
ENSG00000273443      40      43      54      26      48      11     21
ENSG00000217801      60      36     164      62      61      19     34
ENSG00000237330       0       1       0       0       1       1      0
ENSG00000223823       0       0       0       0       0       0      0
                NAPT91 NA3622 NA3659 NA4973 NA18358
ENSG00000188976      1      1      1      0       0
ENSG00000188157      7      7      8     27       6
ENSG00000273443      2      3     78     59      16
ENSG00000217801      8     19    139     68      31
ENSG00000237330      0      0      2      1       0
ENSG00000223823      0      0      0      0       0
# Load colors 

colors <- colorRampPalette(c(brewer.pal(9, "Blues")[1],brewer.pal(9, "Blues")[9]))(100)

pal <- c(brewer.pal(9, "Set1"), brewer.pal(8, "Set2"), brewer.pal(12, "Set3"))
labels <- paste(metaData$Species,metaData$Line, sep=" ")
#PCA function (original code from Julien Roux)
#Load in the plot_scores function
plot_scores <- function(pca, scores, n, m, cols, points=F, pchs =20, legend=F){
  xmin <- min(scores[,n]) - (max(scores[,n]) - min(scores[,n]))*0.05
  if (legend == T){ ## let some room (35%) for a legend                                                                                                                                                 
    xmax <- max(scores[,n]) + (max(scores[,n]) - min(scores[,n]))*0.50
  }
  else {
    xmax <- max(scores[,n]) + (max(scores[,n]) - min(scores[,n]))*0.05
  }
  ymin <- min(scores[,m]) - (max(scores[,m]) - min(scores[,m]))*0.05
  ymax <- max(scores[,m]) + (max(scores[,m]) - min(scores[,m]))*0.05
  plot(scores[,n], scores[,m], xlab=paste("PC", n, ": ", round(summary(pca)$importance[2,n],3)*100, "% variance explained", sep=""), ylab=paste("PC", m, ": ", round(summary(pca)$importance[2,m],3)*100, "% variance explained", sep=""), xlim=c(xmin, xmax), ylim=c(ymin, ymax), type="n")
  if (points == F){
    text(scores[,n],scores[,m], rownames(scores), col=cols, cex=1)
  }
  else {
    points(scores[,n],scores[,m], col=cols, pch=pchs, cex=1.3)
  }
}
# Clustering (original code from Julien Roux)
cors <- cor(counts_genes, method="spearman", use="pairwise.complete.obs")


heatmap.2( cors, scale="none", col = colors, margins = c(12, 12), trace='none', denscol="white", labCol=labels, ColSideColors=pal[as.integer(as.factor(metaData$Species))], RowSideColors=pal[as.integer(as.factor(metaData$Collection))+9], cexCol = 0.2 + 1/log10(15), cexRow = 0.2 + 1/log10(15))

Version Author Date
32b435b brimittleman 2019-11-12
select <- counts_genes
summary(apply(select, 1, var) == 0) 
   Mode   FALSE    TRUE 
logical   32864   11261 
# Perform PCA

pca_genes <- prcomp(t(counts_genes), scale = F)
scores <- pca_genes$x


#Make PCA plots with the factors colored by species

### PCs 1 and 2 Raw Data
for (n in 1:1){
  col.v <- pal[as.integer(metaData$Species)]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
32b435b brimittleman 2019-11-12
### PCs 3 and 4 Raw Data

for (n in 3:3){
  col.v <- pal[as.integer(metaData$Species)]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
32b435b brimittleman 2019-11-12

Plot density for raw data:

density_plot_18504 <- ggplot(counts_genes, aes(x = NA18504)) + geom_density() + labs(title = "Density plot of raw gene counts of NA18504") + labs(x = "Raw counts for each gene")

density_plot_18504

Version Author Date
32b435b brimittleman 2019-11-12

Convert to log2

log_counts_genes <- as.data.frame(log2(counts_genes))
head(log_counts_genes)
                 NA18504  NA18510  NA18523  NA18498  NA18499  NA18502
ENSG00000188976 4.584963 5.643856 4.954196 6.108524 5.087463 5.857981
ENSG00000188157 6.727920 6.022368 6.727920 3.584963 7.000000 5.285402
ENSG00000273443 5.321928 5.426265 5.754888 4.700440 5.584963 3.459432
ENSG00000217801 5.906891 5.169925 7.357552 5.954196 5.930737 4.247928
ENSG00000237330     -Inf 0.000000     -Inf     -Inf 0.000000 0.000000
ENSG00000223823     -Inf     -Inf     -Inf     -Inf     -Inf     -Inf
                  NAPT30   NAPT91   NA3622   NA3659   NA4973  NA18358
ENSG00000188976 1.000000 0.000000 0.000000 0.000000     -Inf     -Inf
ENSG00000188157 2.321928 2.807355 2.807355 3.000000 4.754888 2.584963
ENSG00000273443 4.392317 1.000000 1.584963 6.285402 5.882643 4.000000
ENSG00000217801 5.087463 3.000000 4.247928 7.118941 6.087463 4.954196
ENSG00000237330     -Inf     -Inf     -Inf 1.000000 0.000000     -Inf
ENSG00000223823     -Inf     -Inf     -Inf     -Inf     -Inf     -Inf
density_plot_18504 <- ggplot(log_counts_genes, aes(x = 18504)) + geom_density()

density_plot_18504 + labs(title = "Density plot of log2 counts of 18504") + labs(x = "Log2 counts for each gene") + geom_vline(xintercept = 1) 

Version Author Date
32b435b brimittleman 2019-11-12
plotDensities(log_counts_genes, col=pal[as.numeric(metaData$Species)], legend="topright")

Version Author Date
32b435b brimittleman 2019-11-12

Convert to CPM

cpm <- cpm(counts_genes, log=TRUE)
head(cpm)
                   NA18504   NA18510   NA18523    NA18498   NA18499
ENSG00000188976  0.9877925  1.619259  1.208519  1.9826994  1.067390
ENSG00000188157  3.0590448  1.984237  2.926267 -0.3399951  2.915674
ENSG00000273443  1.6881262  1.411139  1.975775  0.6477686  1.539712
ENSG00000217801  2.2544180  1.167851  3.547531  1.8334849  1.872281
ENSG00000237330 -3.0080795 -2.448042 -3.008080 -3.0080795 -2.455753
ENSG00000223823 -3.0080795 -3.008080 -3.008080 -3.0080795 -3.008080
                   NA18502     NAPT30     NAPT91     NA3622     NA3659
ENSG00000188976  1.8416633 -2.0220152 -2.4030724 -2.4332254 -2.4799651
ENSG00000188157  1.2932560 -1.2206741 -0.7918098 -0.8618518 -0.8265488
ENSG00000273443 -0.3575383  0.4899323 -1.9781161 -1.7043978  2.1408291
ENSG00000217801  0.3308185  1.1354086 -0.6385042  0.3566344  2.9564118
ENSG00000237330 -2.4425685 -3.0080795 -3.0080795 -3.0080795 -2.0942143
ENSG00000223823 -3.0080795 -3.0080795 -3.0080795 -3.0080795 -3.0080795
                    NA4973    NA18358
ENSG00000188976 -3.0080795 -3.0080795
ENSG00000188157  0.7692807 -0.8079148
ENSG00000273443  1.8388114  0.3962125
ENSG00000217801  2.0369808  1.2829205
ENSG00000237330 -2.4514663 -3.0080795
ENSG00000223823 -3.0080795 -3.0080795
plotDensities(cpm, col=pal[as.numeric(metaData$Species)], legend="topright")

Version Author Date
32b435b brimittleman 2019-11-12

Log2 CPM

TMM/log2(CPM)

## Create edgeR object (dge) to calculate TMM normalization  
dge_original <- DGEList(counts=as.matrix(counts_genes), genes=rownames(counts_genes), group = as.character(t(labels)))
dge_original <- calcNormFactors(dge_original)

tmm_cpm <- cpm(dge_original, normalized.lib.sizes=TRUE, log=TRUE, prior.count = 0.25)
head(cpm)
                   NA18504   NA18510   NA18523    NA18498   NA18499
ENSG00000188976  0.9877925  1.619259  1.208519  1.9826994  1.067390
ENSG00000188157  3.0590448  1.984237  2.926267 -0.3399951  2.915674
ENSG00000273443  1.6881262  1.411139  1.975775  0.6477686  1.539712
ENSG00000217801  2.2544180  1.167851  3.547531  1.8334849  1.872281
ENSG00000237330 -3.0080795 -2.448042 -3.008080 -3.0080795 -2.455753
ENSG00000223823 -3.0080795 -3.008080 -3.008080 -3.0080795 -3.008080
                   NA18502     NAPT30     NAPT91     NA3622     NA3659
ENSG00000188976  1.8416633 -2.0220152 -2.4030724 -2.4332254 -2.4799651
ENSG00000188157  1.2932560 -1.2206741 -0.7918098 -0.8618518 -0.8265488
ENSG00000273443 -0.3575383  0.4899323 -1.9781161 -1.7043978  2.1408291
ENSG00000217801  0.3308185  1.1354086 -0.6385042  0.3566344  2.9564118
ENSG00000237330 -2.4425685 -3.0080795 -3.0080795 -3.0080795 -2.0942143
ENSG00000223823 -3.0080795 -3.0080795 -3.0080795 -3.0080795 -3.0080795
                    NA4973    NA18358
ENSG00000188976 -3.0080795 -3.0080795
ENSG00000188157  0.7692807 -0.8079148
ENSG00000273443  1.8388114  0.3962125
ENSG00000217801  2.0369808  1.2829205
ENSG00000237330 -2.4514663 -3.0080795
ENSG00000223823 -3.0080795 -3.0080795
pca_genes <- prcomp(t(tmm_cpm), scale = F)
scores <- pca_genes$x

for (n in 1:2){
  col.v <- pal[as.integer(metaData$Species)]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12

Version Author Date
da4bab0 brimittleman 2019-11-12
# Plot library size

boxplot_library_size <- ggplot(dge_original$samples, aes(x=metaData$Species, y = dge_original$samples$lib.size, fill = metaData$Species)) + geom_boxplot()
 
boxplot_library_size + labs(title = "Library size by Species") + labs(y = "Library size") + labs(x = "Species") + guides(fill=guide_legend(title="Species"))

Version Author Date
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12
plotDensities(tmm_cpm, col=pal[as.numeric(metaData$Species)], legend="topright")

Version Author Date
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12

Filter low expressed gene

Filter based on log2 cpm

filter log2(cpm >1) in at least 10 of the samples (2/3)

#filter counts
keep.exprs=rowSums(tmm_cpm>1) >8

counts_filtered= counts_genes[keep.exprs,]




plotDensities(counts_filtered, col=pal[as.numeric(metaData$Species)], legend="topright")

Version Author Date
da4bab0 brimittleman 2019-11-12
labels <- paste(metaData$Species, metaData$Line, sep=" ")
dge_in_cutoff <- DGEList(counts=as.matrix(counts_filtered), genes=rownames(counts_filtered), group = as.character(t(labels)))
dge_in_cutoff <- calcNormFactors(dge_in_cutoff)

cpm_in_cutoff <- cpm(dge_in_cutoff, normalized.lib.sizes=TRUE, log=TRUE, prior.count = 0.25)
head(cpm_in_cutoff)
                 NA18504    NA18510  NA18523  NA18498  NA18499   NA18502
ENSG00000217801 2.248146  1.0811901 3.657635 1.792969 1.843394 0.2076129
ENSG00000186891 5.023562  5.0327842 4.883988 6.663573 5.691238 3.3594991
ENSG00000186827 3.107235  4.6526859 1.697751 1.815945 3.957501 1.8998264
ENSG00000078808 6.873762  7.0290328 7.460760 7.026971 6.609430 6.7348224
ENSG00000176022 4.767653  4.6792021 4.786920 5.258909 4.597113 4.6984962
ENSG00000184163 1.432462 -0.1665159 1.627720 1.395504 1.046753 0.3500809
                  NAPT30     NAPT91   NA3622   NA3659   NA4973  NA18358
ENSG00000217801 1.164822 -0.7903107 0.279009 3.001252 2.105148 1.303916
ENSG00000186891 4.637726  3.1601319 4.730313 6.450470 6.482165 6.485174
ENSG00000186827 0.256488  2.0075331 3.153929 0.649213 5.258424 2.178300
ENSG00000078808 6.691955  6.7661750 6.779328 6.786500 7.215767 7.064091
ENSG00000176022 4.835934  5.5508794 4.848618 5.278743 4.794461 5.564750
ENSG00000184163 2.902942  2.1962979 1.412850 1.768892 1.691914 3.225391
hist(cpm_in_cutoff, xlab = "Log2(CPM)", main = "Log2(CPM) values for genes meeting the filtering criteria", breaks = 100 )

Version Author Date
da4bab0 brimittleman 2019-11-12

Voom transformation:

Species <- factor(metaData$Species)
design <- model.matrix(~ 0 + Species)
head(design)
  SpeciesChimp SpeciesHuman
1            1            0
2            1            0
3            1            0
4            1            0
5            1            0
6            1            0
colnames(design) <- gsub("Species", "", dput(colnames(design)))
c("SpeciesChimp", "SpeciesHuman")

Voom creates a random effect.

# Voom with individual as a random variable

cpm.voom<- voom(counts_filtered, design, normalize.method="quantile", plot=T)

Version Author Date
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12
boxplot(cpm.voom$E, col = pal[as.numeric(metaData$Species)],las=2)

Version Author Date
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12
plotDensities(cpm.voom, col =  pal[as.numeric(metaData$Species)], legend = "topleft") 

Version Author Date
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12

Looks like i still have a skew on the lower side of the distribution.

# PCA 

pca_genes <- prcomp(t(cpm.voom$E), scale = T)
scores <- pca_genes$x


eigsGene <- pca_genes$sdev^2
proportionG = eigsGene/sum(eigsGene)

plot(proportionG)

Version Author Date
a22bae9 brimittleman 2019-11-13
da4bab0 brimittleman 2019-11-12
32b435b brimittleman 2019-11-12
for (n in 1:2){
  col.v <- pal[as.integer(metaData$Species)]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13
da4bab0 brimittleman 2019-11-12

Version Author Date
a22bae9 brimittleman 2019-11-13
#Clustering (original code from Julien Roux)
cors <- cor(cpm.voom$E, method="spearman", use="pairwise.complete.obs")


heatmap.2( cors, scale="none", col = colors, margins = c(12, 12), trace='none', denscol="white", labCol=labels, ColSideColors=pal[as.integer(as.factor(metaData$Species))], RowSideColors=pal[as.integer(as.factor(metaData$Species))+9], cexCol = 0.2 + 1/log10(15), cexRow = 0.2 + 1/log10(15))

Version Author Date
da4bab0 brimittleman 2019-11-12

This is wierd. Normalization moves 2 samples to opposite species clusters but the samples that separate in the correlation are not those samples. 4973 and 18498 are the samples that looked funny on the original 3’ data. This may be a sample swap at the RNA stage. These samples were in the same extraction batch. It could have happened then. I will look into this more.

One thing I can do is look at the correlation between the PCs and other factors in the data.

# PCA 

pca_genes <- prcomp(t(cpm.voom$E), scale = F)
scores <- pca_genes$x

for (n in 1:2){
  col.v <- pal[as.integer(metaData$Collection)]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13

Version Author Date
a22bae9 brimittleman 2019-11-13
metaData$Extraction=as.factor(metaData$Extraction)

for (n in 1:2){
  col.v <- pal[as.integer(metaData$Extraction)]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13

Version Author Date
a22bae9 brimittleman 2019-11-13

It does not look like batch (who collected or extraction date batch)

cols = brewer.pal(9, "Blues")
palC = colorRampPalette(cols)


metaData$UndilutedAverageorder = findInterval(metaData$UndilutedAverage, sort(metaData$UndilutedAverage))
for (n in 1:2){
  col.v <- palC(nrow(metaData))[metaData$UndilutedAverageorder]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13

Version Author Date
a22bae9 brimittleman 2019-11-13
metaData$BioAConcorder = findInterval(metaData$BioAConc, sort(metaData$BioAConc))
for (n in 1:2){
  col.v <- palC(nrow(metaData))[metaData$BioAConcorder]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13

Version Author Date
a22bae9 brimittleman 2019-11-13
metaData$RinConcorder = findInterval(metaData$Rin, sort(metaData$Rin))
for (n in 1:2){
  col.v <- palC(nrow(metaData))[metaData$RinConcorder]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13

Version Author Date
a22bae9 brimittleman 2019-11-13

The samples do not cluster by collection concentration, RNA rin score or RNA concentration.

metaData$AssignedOrthoorder = findInterval(metaData$AssignedOrtho, sort(metaData$AssignedOrtho))
for (n in 1:2){
  col.v <- palC(nrow(metaData))[metaData$AssignedOrthoorder]
  plot_scores(pca_genes, scores, n, n+1, col.v)
}

Version Author Date
a22bae9 brimittleman 2019-11-13

Version Author Date
a22bae9 brimittleman 2019-11-13

They also do not cluster by number of reads mapping to ortho exons.

PCA heatmap: Code from Michelle Ward:

x.pca <- pca_genes

tech_factors <- metaData
tech_factors_sum <- tech_factors[,c(2:14)] %>% select(-CollectionDate)

p_comps <- 1:6
pc_cov_cor <- matrix(nrow = ncol(tech_factors_sum), ncol = length(p_comps),
                     dimnames = list(colnames(tech_factors_sum), colnames(x.pca$x)[p_comps]))
for (pc in p_comps) {
  for (covariate in 1:ncol(tech_factors_sum)) {
    lm_result <- lm(x.pca$x[, pc] ~ tech_factors_sum[, covariate])
    r2 <- summary(lm_result)$r.squared
    pc_cov_cor[covariate, pc] <- r2
  }
}

pc_cov_pval <- matrix(nrow = ncol(tech_factors_sum), ncol = length(p_comps),
                      dimnames = list(colnames(tech_factors_sum), colnames(x.pca$x)[p_comps]))

for (pc in p_comps) {
  for (covariate_2 in 1:ncol(tech_factors_sum)) {
    lm_result_2 <- lm(x.pca$x[, pc] ~ tech_factors_sum[, covariate_2])
    pval <- anova(lm_result_2)$'Pr(>F)'[1]
    pc_cov_pval[covariate_2, pc] <- pval
  }
}

PCs <- c("PC1", "PC2", "PC3", "PC4", "PC5", "PC6")
Tech_fac <- colnames(tech_factors_sum)
#Tech_fac <- c("Species",   "Individual", "O2.",  "Condition" , "Sex", "RIN" , "CO2", "Purity_high", "Purity_med" ,
              #"Expt_Batch", "RNA_Batch", "Library_Batch", "Seq_pool", "Episomal_integration" )

heatmap.2(as.matrix(pc_cov_cor[Tech_fac,PCs]),col=brewer.pal(4, "Greens"), trace="none",
          Rowv=FALSE, Colv=FALSE, key=T, main="Cor. PCs & tech factors", dendrogram="none",
          key.title=NA, cexRow=0.9, cexCol=0.9)

Version Author Date
a22bae9 brimittleman 2019-11-13
log10_pc_cov_pval <- -log(pc_cov_pval)
heatmap.2(as.matrix(log10_pc_cov_pval[Tech_fac,PCs]), col=brewer.pal(9, "Greens"), trace="none",
          Rowv=FALSE, Colv=FALSE, key=T, main="-log10 pval of cor. PCs & tech factors", dendrogram="none",
          key.title=NA, cexRow=0.9, cexCol=0.9)

Test for DE

fit.cpm.voom = lmFit(cpm.voom, design, plot=T)
head(coef(fit.cpm.voom))
                    Chimp    Human
ENSG00000217801 1.8176205 1.053311
ENSG00000186891 5.1291883 5.317657
ENSG00000186827 2.9252639 2.080906
ENSG00000078808 6.9485052 6.883254
ENSG00000176022 4.7991489 5.131756
ENSG00000184163 0.8725114 2.121651
contr <- makeContrasts(Chimp - Human, levels = colnames(coef(fit.cpm.voom)))
contr
       Contrasts
Levels  Chimp - Human
  Chimp             1
  Human            -1
tmp <- contrasts.fit(fit.cpm.voom, contr)
tmp <- eBayes(tmp)
top.table <- topTable(tmp, sort.by = "P", n = Inf)
head(top.table, 20)
                    logFC  AveExpr         t      P.Value    adj.P.Val
ENSG00000205531 -5.906048 5.110707 -42.42147 1.003413e-14 6.098694e-11
ENSG00000204463 -6.197073 5.328838 -41.83367 1.190221e-14 6.098694e-11
ENSG00000133112  6.023976 8.508367  37.04918 5.253426e-14 1.794570e-10
ENSG00000186298  3.188104 6.854108  33.66982 1.687825e-13 4.324209e-10
ENSG00000105372  4.894937 7.897387  31.55050 3.726742e-13 7.638331e-10
ENSG00000142937  5.663078 8.145949  30.69857 5.200238e-13 8.882006e-10
ENSG00000145741  6.238790 5.598248  28.94136 1.064803e-12 1.558871e-09
ENSG00000147604 -5.969523 5.605022 -28.56893 1.246222e-12 1.596410e-09
ENSG00000071082  5.667843 6.554807  26.24239 3.491956e-12 3.874800e-09
ENSG00000100316  5.948079 8.194419  26.07064 3.781030e-12 3.874800e-09
ENSG00000072864 -6.601953 4.740080 -24.35154 8.629478e-12 8.039536e-09
ENSG00000142541  6.210689 6.651181  23.67698 1.211455e-11 1.034583e-08
ENSG00000116478  3.311414 5.888231  23.30107 1.469479e-11 1.158402e-08
ENSG00000131475  3.194285 3.623143  20.35316 7.473461e-11 5.341737e-08
ENSG00000105640 -2.988556 8.173442 -20.27660 7.818702e-11 5.341737e-08
ENSG00000136021  3.428259 4.629566  20.00332 9.198528e-11 5.891657e-08
ENSG00000169926 -5.926609 5.806506 -19.38256 1.341196e-10 8.085047e-08
ENSG00000198034  6.187416 7.128275  19.23026 1.473766e-10 8.390641e-08
ENSG00000154978 -2.562136 5.114129 -19.00708 1.694215e-10 9.016813e-08
ENSG00000171863  3.241360 6.619888  18.94677 1.759721e-10 9.016813e-08
                       B
ENSG00000205531 21.38525
ENSG00000204463 21.34027
ENSG00000133112 22.07762
ENSG00000186298 21.10822
ENSG00000105372 20.42489
ENSG00000142937 20.10921
ENSG00000145741 18.59467
ENSG00000147604 18.55697
ENSG00000071082 18.09236
ENSG00000100316 18.30545
ENSG00000072864 16.62325
ENSG00000142541 16.96488
ENSG00000116478 16.94969
ENSG00000131475 15.01223
ENSG00000105640 15.42738
ENSG00000136021 15.04792
ENSG00000169926 14.70272
ENSG00000198034 14.74804
ENSG00000154978 14.60662
ENSG00000171863 14.61913
length(which(top.table$adj.P.Val < 0.05))
[1] 449

Make a table to plot:

-log10(bh adjusted pval) vs logFC (log3 fold change)

top.table=top.table %>% mutate(Species=ifelse(logFC > 1 & adj.P.Val<.05, "Chimp", ifelse(logFC < -1 & adj.P.Val< .05, "Human", "Neither")))
  
  

ggplot(top.table, aes(x=logFC, y= -log10(adj.P.Val))) + geom_point(aes(col=Species), alpha=.3)

summary(decideTests(tmp))
       Chimp - Human
Down             191
NotSig          9799
Up               258

questions: * Correlation in fit to use design rather than contrast matrix. * direction of effect size * what is eBayes step?

Rerun without bad samples.


sessionInfo()
R version 3.5.1 (2018-07-02)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Scientific Linux 7.4 (Nitrogen)

Matrix products: default
BLAS/LAPACK: /software/openblas-0.2.19-el7-x86_64/lib/libopenblas_haswellp-r0.2.19.so

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] grid      stats     graphics  grDevices utils     datasets  methods  
[8] base     

other attached packages:
 [1] reshape2_1.4.3      RColorBrewer_1.1-2  VennDiagram_1.6.20 
 [4] futile.logger_1.4.3 R.utils_2.7.0       R.oo_1.22.0        
 [7] R.methodsS3_1.7.1   edgeR_3.24.0        limma_3.38.2       
[10] gplots_3.0.1        scales_1.0.0        forcats_0.3.0      
[13] stringr_1.3.1       dplyr_0.8.0.1       purrr_0.3.2        
[16] readr_1.3.1         tidyr_0.8.3         tibble_2.1.1       
[19] ggplot2_3.1.1       tidyverse_1.2.1     workflowr_1.5.0    

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.2           locfit_1.5-9.1       lubridate_1.7.4     
 [4] lattice_0.20-38      gtools_3.8.1         assertthat_0.2.0    
 [7] rprojroot_1.3-2      digest_0.6.18        R6_2.3.0            
[10] cellranger_1.1.0     plyr_1.8.4           futile.options_1.0.1
[13] backports_1.1.2      evaluate_0.12        httr_1.3.1          
[16] pillar_1.3.1         rlang_0.4.0          lazyeval_0.2.1      
[19] readxl_1.1.0         rstudioapi_0.10      gdata_2.18.0        
[22] whisker_0.3-2        rmarkdown_1.10       labeling_0.3        
[25] munsell_0.5.0        broom_0.5.1          compiler_3.5.1      
[28] httpuv_1.4.5         modelr_0.1.2         pkgconfig_2.0.2     
[31] htmltools_0.3.6      tidyselect_0.2.5     crayon_1.3.4        
[34] withr_2.1.2          later_0.7.5          bitops_1.0-6        
[37] nlme_3.1-137         jsonlite_1.6         gtable_0.2.0        
[40] formatR_1.5          git2r_0.26.1         magrittr_1.5        
[43] KernSmooth_2.23-15   cli_1.1.0            stringi_1.2.4       
[46] fs_1.3.1             promises_1.0.1       xml2_1.2.0          
[49] generics_0.0.2       lambda.r_1.2.3       tools_3.5.1         
[52] glue_1.3.0           hms_0.4.2            yaml_2.2.0          
[55] colorspace_1.3-2     caTools_1.17.1.1     rvest_0.3.2         
[58] knitr_1.20           haven_1.1.2