Question

ReportingTools include pvalue NA and comparissons

0

Entering edit mode

NMostajo ▴ 10

@nmostajo-10900

Last seen 8.4 years ago

Germany

Dear all,

I am trying to analyze my smallRNAseq data with Deseq2 and to run ReportingTools on it.

I have different treatments and timepoints, I saw that you have a coef option to tell where to do the comparison with, but my data set then will exclude another timepoint or treatment. I am wondering if it is possible to get all pairwise comparisons from the data set as log2FC, or to input the results from Deseq2 into those columns and keep the barplots.

I tried to do them first as pairwise comparisons, but that did not seem to work properly:

dds.sub1 <- dds[ , dds$condition %in% c("M_3h", "R_3h") ]
dds.sub1$condition <- droplevels(dds.sub1$condition)
dds.sub1$type <- droplevels(dds.sub1$type)

publish(dds.sub1,des2Report, pvalueCutoff=1.1, annotation.db="org.Hs.egENSEMBL", factor = colData(dds.sub1)$condition, reportDir=out, n = length(row.names(dds.sub1)))

Error in counts(object) %*% whichSamples : non-conformable arguments

In case you have a comment or suggestion on that it would be nice.

I am also wondering if there is a way to print everything, even if the calculated pvalue is NA, I have one replicate that I would not exclude, but caused that some interesting genes get a pvalue of NA, and for me would be interesting to check them.

Thank you!

R version 3.2.4 (2016-03-10)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Debian GNU/Linux 8 (jessie)

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

attached base packages:
[1] parallel  stats4    stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] ReportingTools_2.10.0      knitr_1.12.3               pheatmap_1.0.8             stringr_1.0.0              RColorBrewer_1.1-2        
 [6] gplots_3.0.1               DESeq2_1.10.1              RcppArmadillo_0.6.600.4.0  Rcpp_0.12.4                SummarizedExperiment_1.0.2
[11] GenomicRanges_1.22.4       GenomeInfoDb_1.6.3         ggplot2_2.1.0              geneplotter_1.48.0         annotate_1.48.0           
[16] XML_3.98-1.4               lattice_0.20-33            genefilter_1.52.1          org.Hs.eg.db_3.2.3         AnnotationDbi_1.32.3      
[21] IRanges_2.4.8              S4Vectors_0.8.11           Biobase_2.30.0             BiocGenerics_0.16.1        scatterplot3d_0.3-37      
[26] RSQLite_1.0.0              DBI_0.3.1                 

loaded via a namespace (and not attached):
 [1] edgeR_3.12.1             splines_3.2.4            R.utils_2.3.0            gtools_3.5.0             Formula_1.2-1           
 [6] latticeExtra_0.6-28      RBGL_1.46.0              BSgenome_1.38.0          Rsamtools_1.22.0         Category_2.36.0         
[11] biovizBase_1.18.0        limma_3.26.9             XVector_0.10.0           colorspace_1.2-6         ggbio_1.18.5            
[16] R.oo_1.20.0              Matrix_1.2-5             plyr_1.8.3               OrganismDbi_1.12.1       GSEABase_1.32.0         
[21] biomaRt_2.26.1           zlibbioc_1.16.0          xtable_1.8-2             GO.db_3.2.2              scales_0.4.0            
[26] gdata_2.17.0             BiocParallel_1.4.3       PFAM.db_3.2.2            GenomicFeatures_1.22.13  nnet_7.3-12             
[31] survival_2.39-2          magrittr_1.5             R.methodsS3_1.7.1        GGally_1.0.1             hwriter_1.3.2           
[36] foreign_0.8-66           GOstats_2.36.0           graph_1.48.0             BiocInstaller_1.20.3     tools_3.2.4             
[41] munsell_0.4.3            locfit_1.5-9.1           cluster_2.0.4            lambda.r_1.1.7           Biostrings_2.38.4       
[46] caTools_1.17.1           futile.logger_1.4.1      grid_3.2.4               RCurl_1.95-4.8           dichromat_2.0-0         
[51] VariantAnnotation_1.16.4 AnnotationForge_1.12.2   bitops_1.0-6             gtable_0.2.0             reshape_0.8.5           
[56] reshape2_1.4.1           GenomicAlignments_1.6.3  gridExtra_2.2.1          rtracklayer_1.30.4       Hmisc_3.17-3            
[61] futile.options_1.0.0     KernSmooth_2.23-15       stringi_1.0-1            rpart_4.1-10             acepack_1.3-3.3

R reportingtools deseq2 pvalue coef • 1.9k views

ADD COMMENT • link updated 8.4 years ago by Michael Love 42k • written 8.4 years ago by NMostajo ▴ 10

score 0 · Answer 1 · 2016-06-14

0

Entering edit mode

Michael Love 42k

@mikelove

Last seen 14 hours ago

United States

I'd strongly recommend to build the results table outside of ReportingTools, so you have full control through the DESeq2 results() function.

You can then pass the results table and the dds object to ReportingTools.

The ReportingTools authors and I worked out this solution for just this kind of situation.

As far as NA p-values, you can get rid of them if you need to on the DESeq2 side using cooksCutoff=FALSE. You could inspect indiviudal cases using plotCounts (see vignette) to see if the outlier detection is needed or not.

ADD COMMENT • link 8.4 years ago Michael Love 42k

0

Entering edit mode

That means that I should run results for the overall dds (until now I only did contrast analysis)? So I do not loose this NA values and run the ReportingTools over this OverallComparisson. If I understood correctly? Then I would have to??

OverallComparisson <- results(dds, cooksCutoff = FALSE)

> OverallComparisson
Error in .classEnv(classDef) : 
  trying to get slot "package" from an object of a basic class ("NULL") with no slots

interesting because:

> str(OverallComparisson)
Formal class 'DESeqResults' [package "DESeq2"] with 6 slots
  ..@ rownames       : chr [1:41516] "ENSG00000223972" "ENSG00000227232" "ENSG00000278267" "ENSG00000243485" ...
  ..@ nrows          : int 41516
  ..@ listData       :List of 6
  .. ..$ baseMean      : num [1:41516] 0.975 5.018 12.088 2.884 0.126 ...
  .. ..$ log2FoldChange: num [1:41516] 0.3969 0.2085 -0.4072 0.9975 0.0482 ...
  .. ..$ lfcSE         : num [1:41516] 0.369 0.589 0.439 0.656 0.317 ...
  .. ..$ stat          : num [1:41516] 1.075 0.354 -0.928 1.52 0.152 ...
  .. ..$ pvalue        : num [1:41516] 0.282 0.723 0.353 0.128 0.879 ...
  .. ..$ padj          : num [1:41516] NA 0.986 0.895 0.666 NA ...
  ..@ elementType    : chr "ANY"
  ..@ elementMetadata:Formal class 'DataFrame' [package "S4Vectors"] with 6 slots
  .. .. ..@ rownames       : NULL
  .. .. ..@ nrows          : int 6
  .. .. ..@ listData       :List of 2

Also in padj there are NA values.

I do not seem to understand the problem, or what did I do wrong.

### I know this will fail, but still to be sure I understoo thisOvCo <- DESeqResults(OverallComparisson)

publish(OverallComparisson,des2Report, pvalueCutoff=1.1, annotation.db="org.Hs.egENSEMBL", factor = colData(dds)$condition, reportDir=out, n = length(row.names(dds)))

Error in .make.gene.plots(df, eSet, factor, figure.directory, ...) : 
  No expression data was provided. Nothing to plot.

ADD REPLY • link 8.4 years ago NMostajo ▴ 10

0

Entering edit mode

You can use whatever contrasts you like, I'm just suggesting that you pass a results object to ReportingTools.

I'm a bit lost and can't figure out what exactly you want from ReportingTools and what code you've run already.

Can you describe in detail what you want the report to look like?

ADD REPLY • link 8.4 years ago Michael Love 42k

0

Entering edit mode

Sorry, I have run already ReportingTools, and what I get (and I would like to add)

Gene names - plots - FC (FC for all comparissons**) : Including the genes with a padj of NA

**I understand that the FC of each pairwise comparisons has to be added later

My problem is with the padj NA. According to the manual I do not need to put a contrast so I can do

>OverallComparisson <- results(dds, cooksCutoff = FALSE)

which runs, but when I want to analyze it I get :

> OverallComparisson

Error in .classEnv(classDef) : 
  trying to get slot "package" from an object of a basic class ("NULL") with no slots

and ReportingTools does not recognize it as an input, it is the same class as dds and any of my contrasts, but it can not read it. I assume is because R is also not recognizing the output with cooksC = Null.

what works:

>R24hZ24h <- results(dds,contrast = c("condition","Z_24h","R_24h"))
>summary(R24hZ24h )

> R24hZ24h
log2 fold change (MAP): condition Z_24h vs R_24h 
Wald test p-value: condition Z_24h vs R_24h 
DataFrame with 41516 rows and 6 columns
                  baseMean log2FoldChange     lfcSE       stat       pvalue        padj
                 <numeric>      <numeric> <numeric>  <numeric>    <numeric>   <numeric>
ENSG00000223972  0.9754851      0.1424935 0.3686457  0.3865324           NA          NA
ENSG00000227232  5.0175895      0.3990592 0.5604216  0.7120695    0.4764217   0.8382602
ENSG00000278267 12.0882809      0.1337271 0.4091992  0.3268020    0.7438177   0.9349837
ENSG00000243485  2.8841597      0.5553056 0.6587982  0.8429069    0.3992805          NA

My full command for ReportingTools (this was the first try with a Pvalue restriction):

out <- "/folder/"
des2Report05 <- HTMLReport(shortName = 'RNAseq_analysis_with_DESeq2_p05', title = 'sRNA-seq THP1 analysis using DESeq2, pvalue cutoff 0.05', basePath = out, reportDirectory = 'html/')

publish(dds,des2Report05, pvalueCutoff=0.05, annotation.db="org.Hs.egENSEMBL", factor = colData(dds)$condition, reportDir=out, n = length(row.names(dds)))

finish(des2Report05)

In genral I can only put as input to publish the dds, but not the results even when I try to transform them.

Sorry if I do not make sense, I do not know which other information I should add.

ADD REPLY • link 8.4 years ago NMostajo ▴ 10

0

Entering edit mode

Adjusted p-values of NA are from independent filtering. You can turn this off with independentFiltering=FALSE, or you can replace NA with 1. See the DESeq2 vignette on how to do this.

I guess getting exactly what you want from ReportingTools might require input from those package authors on this thread.

I don't know why you are getting the "Error in .classEnv(classDef)" error, but I can't see what code you've run before this line, so I can't help much.

What happens in a fresh R session if you: build the DESeqDataSet, run DESeq(), then run results(dds, cooksCutoff=FALSE)? Do you still get this error?

ADD REPLY • link 8.4 years ago Michael Love 42k

0

Entering edit mode

I tried from the beginning and I got the same result:

dds <- DESeqDataSetFromMatrix(countData = THP1_countData,
colData = colData_try,
design = ~ condition)

featureData <- data.frame(gene=rownames(THP1_countData))

dds <- dds[ rowSums(counts(dds)) > 1, ]

dds$condition <- factor(dds$condition, levels=c("M_3h","M_24h", "R_3h" ,"R_24h" ,"Z_3h","Z_24h"))
dds <- DESeq(dds)

OverallComparisson <- results(dds, cooksCutoff = FALSE)
R3hZ3hNA <- results(dds,contrast = c("condition","Z_3h","R_3h"), cooksCutoff = FALSE)
summary(R3hZ3hNA)

out of 41516 with nonzero total read count
adjusted p-value < 0.1
LFC > 0 (up)     : 185, 0.45% 
LFC < 0 (down)   : 388, 0.93% 
outliers [1]     : 0, 0% 
low counts [2]   : 34610, 83% 
(mean count < 10)
[1] see 'cooksCutoff' argument of ?results
[2] see 'independentFiltering' argument of ?results

M3hR3h <- results(dds,contrast = c("condition","R_3h", "M_3h"))
summary(M3hR3h)

out of 41516 with nonzero total read count
adjusted p-value < 0.1
LFC > 0 (up)     : 223, 0.54% 
LFC < 0 (down)   : 149, 0.36% 
outliers [1]     : 728, 1.8% 
low counts [2]   : 37209, 90% 
(mean count < 19)
[1] see 'cooksCutoff' argument of ?results
[2] see 'independentFiltering' argument of ?results

> R3hZ3hNA
Error in .classEnv(classDef) : 
  trying to get slot "package" from an object of a basic class ("NULL") with no slots
> OverallComparisson
Error in .classEnv(classDef) : 
  trying to get slot "package" from an object of a basic class ("NULL") with no slots

However the structure of all the three analysis is the same.

Thanks for the help! (sorry for the delay on answer, seems as a newbie in the forum I cant answer that many times in a row)

ADD REPLY • link 8.4 years ago NMostajo ▴ 10

0

Entering edit mode

Sorry, I don't know the issue, but it looks to be possibly something with packages not working together properly. But i haven't seen this reported before. You're using an old release of Bioconductor, maybe the error would go away if you updated to the current release.