Hi,

I changed to R version 3.2.0 and DiffBind version 1.14.3. The problems are:

- The heatmap produced by

data = dba(sampleSheet="HistoneData_C.csv", minOverlap=2)
plot(data)

is different that in previous version that I had (1.8.5) (grouping of samples is different). Is this because of a change in the method or a bug? I tried several versions of parameters.

- The side colours in heatmap do not work - see below.

Would be grateful for solution or explanation.

Best

Pawel

R version 3.2.0 (2015-04-16)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

locale:
[1] LC_COLLATE=Polish_Poland.1250  LC_CTYPE=Polish_Poland.1250    LC_MONETARY=Polish_Poland.1250
[4] LC_NUMERIC=C                   LC_TIME=Polish_Poland.1250    

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

other attached packages:
 [1] DiffBind_1.14.3         RSQLite_1.0.0           DBI_0.3.1               locfit_1.5-9.1         
 [5] GenomicAlignments_1.4.1 Rsamtools_1.20.2        Biostrings_2.36.1       XVector_0.8.0          
 [9] limma_3.24.4            GenomicRanges_1.20.3    GenomeInfoDb_1.4.0      IRanges_2.2.1          
[13] S4Vectors_0.6.0         BiocGenerics_0.14.0     BiocInstaller_1.18.2   

loaded via a namespace (and not attached):
 [1] Rcpp_0.11.6            lattice_0.20-31        GO.db_3.1.2            gtools_3.4.2          
 [5] digest_0.6.8           plyr_1.8.2             futile.options_1.0.0   BatchJobs_1.6         
 [9] ShortRead_1.26.0       ggplot2_1.0.1          gplots_2.17.0          zlibbioc_1.14.0       
[13] annotate_1.46.0        gdata_2.16.1           Matrix_1.2-0           checkmate_1.5.3       
[17] systemPipeR_1.2.1      proto_0.3-10           GOstats_2.34.0         splines_3.2.0         
[21] BiocParallel_1.2.1     stringr_1.0.0          pheatmap_1.0.2         munsell_0.4.2         
[25] sendmailR_1.2-1        base64enc_0.1-2        BBmisc_1.9             fail_1.2              
[29] edgeR_3.10.0           XML_3.98-1.1           AnnotationForge_1.10.1 MASS_7.3-40           
[33] bitops_1.0-6           grid_3.2.0             RBGL_1.44.0            xtable_1.7-4          
[37] GSEABase_1.30.1        gtable_0.1.2           magrittr_1.5           scales_0.2.4          
[41] graph_1.46.0           KernSmooth_2.23-14     amap_0.8-14            stringi_0.4-1         
[45] hwriter_1.3.2          reshape2_1.4.1         genefilter_1.50.0      latticeExtra_0.6-26   
[49] futile.logger_1.4.1    brew_1.0-6             rjson_0.2.15           lambda.r_1.1.7        
[53] RColorBrewer_1.1-2     tools_3.2.0            Biobase_2.28.0         Category_2.34.2       
[57] survival_2.38-1        AnnotationDbi_1.30.1   colorspace_1.2-6       caTools_1.17.1        
>

There were several changes to heatmap since 1.8.5

http://www.bioconductor.org/packages/release/bioc/news/DiffBind/NEWS

Easiest solution would probably be to downgrade until task is done unless you want to dig through code. Did you forget to attach error/code for heatmap side colors?

Hi Pawel-

Are the changes relatively minor (order of samples in a cluster) or quite major (completely different clustering)? Perhaps a screenshot of the before and after plots would help me see if this should be happening.

Many things have changed since version 1.8, not only in DiffBind but also in the statistical packages that DiffBind depends on. Any of these could cause a change in cluster behavior, There were some small changes made in the way that DiffBind reads in peak scores and normalizes them, which could also cause changes. 

Could you be more specific regarding your remark that "The side colours in heatmap do not work"? 

Cheers-

Rory

Hi,

Thanks for answering.

- The problem with "side colurs" is that the coloured bar indicating groups of samples in the heatmap does not show properly - it is gray independently of the choice of the grouping variable.

- The differences in clustering (in comparison to previous version) depend on the set of samples analysed. I trust that the changes in the packages are made to improve them so I will try to accept the current version of the algorithm and of results.

Best

Pawel

Hi Pawel-

Could you send me your code for the call to dba,plotHeatmap()? 

-Rory