Scaling DEXSeq estimateDispersions for large datasets, glmGamPoi?
2
1
Entering edit mode
jeremymsimon ▴ 20
@4174dc58
Last seen 19 months ago
United States

Hi,

I'd like to revisit some earlier posts regarding running DEXSeq on a large number of samples, similar to past posts like this one. My eventual goal is to run DEXSeq on a subset of TCGA tumors; my full dataset comprises ~300 samples x ~700k junctions, and I (perhaps expectedly) have run into issues where estimateDispersions(dxd) will run forever, even with BPPARAM=SnowParam(4) specified. The post linked above refers to some improvements as part of the DEXSeqAlt repository, however since its creation 6 years ago, DESeq2::estimateDispersions() can now utilize glmGamPoi to optimize this step to scale more efficiently. I'm wondering if it is possible for DEXSeq to be updated to adopt glmGamPoi and improve the speed of this step / scale better to large datasets, or if I could simply run bits of the DESeq2 + glmGamPoi workflow as described here on my dxd object.

I realize the large dimensions of these data may not be feasible no matter what. Therefore for testing purposes, I've created a smaller-scale example dataset available here where I've randomly selected 10 tumors from each test condition (10 in group "low", 10 in group "high", 20 total samples), and have removed entire genes whose total count was in the bottom 50% (ie restrict to only junctions within genes whose expression is in the top 50%). This cuts down the dimensions of my DEXSeq object to 558591 x 40. A (hopefully) more realistic goal may be to test many of these random subsets as a permutation, but I'm not sure whether pre-selecting genes/junctions to the top 50% of expressers is statistically appropriate given the model.

The original source of my data is from here, and I've gathered the data in a SummarizedExperiment object then converted to a DEXSeqDataSet object with DEXSeqDataSetFromSE(se, design=formula( ~ sample + exon + Group:exon)). The resulting object seems to be consistent with what you describe in the vignette where the first 20 columns represent counts from the given exon (this) and the next 20 columns represent the sum of the other exons belonging to the same gene (others). Happy to share my code for any of these wrangling steps if it would be helpful.

Assuming I've set everything up correctly, you should be able to reproduce and test the speed with something like the following code. For me, the below completed successfully in ~1.5 hrs using 4 cores on our compute cluster.

Of course if I've missed something obvious or there are already documented workarounds somewhere, please let me know!

Thanks so much!

library(DEXSeq)
library(BiocParallel)

# Load random sampling
dxd <- readRDS(file="DEXSeq_Bioc_rand1.rds")

# Peek at coldata
colData(dxd)

DataFrame with 40 rows and 5 columns
      sample         SampleID    Group         sample.1     exon
    <factor>      <character> <factor>         <factor> <factor>
1          1 TCGA-E2-A1IH-01A      low TCGA.E2.A1IH.01A     this
2          2 TCGA-C8-A1HO-01A      low TCGA.C8.A1HO.01A     this
3          3 TCGA-A2-A1FW-01A      low TCGA.A2.A1FW.01A     this
4          4 TCGA-C8-A12U-01A      low TCGA.C8.A12U.01A     this
5          5 TCGA-C8-A273-01A      low TCGA.C8.A273.01A     this
...      ...              ...      ...              ...      ...
36        16 TCGA-BH-A0DH-01A     high TCGA.BH.A0DH.01A   others
37        17 TCGA-AO-A12G-01A     high TCGA.AO.A12G.01A   others
38        18 TCGA-D8-A27T-01A     high TCGA.D8.A27T.01A   others
39        19 TCGA-A8-A094-01A     high TCGA.A8.A094.01A   others
40        20 TCGA-D8-A1XF-01A     high TCGA.D8.A1XF.01A   others

# Peek at counts
counts(dxd)[1:5,1:5]

                        [,1] [,2] [,3] [,4] [,5]
ENSG00000000003.10:E001    0    0    0    0    0
ENSG00000000003.10:E002    0    0    0    0    0
ENSG00000000003.10:E003  259   43  109    6  153
ENSG00000000003.10:E004    0    0    0    0    0
ENSG00000000003.10:E005    0    0    0    0    0


# Now proceed with DEXSeq
dxd <- estimateSizeFactors(dxd)
options(bphost="localhost")
dxd <- estimateDispersions(dxd, BPPARAM=SnowParam(4, log=TRUE) )
dxd <- testForDEU(dxd, BPPARAM=SnowParam(4, log=TRUE) )
dxd <- estimateExonFoldChanges(dxd, fitExpToVar="Group", BPPARAM=SnowParam(4, log=TRUE), denominator="low")
dxr1 <- DEXSeqResults(dxd)


head(as.data.frame(dxr1[!is.na(dxr1$padj) & dxr1$padj < 0.05,]))

                                   groupID featureID exonBaseMean dispersion     stat       pvalue
ENSG00000003402.15:E102 ENSG00000003402.15      E102     5.682785 0.01433858 17.75352 2.514524e-05
ENSG00000006071.7:E054   ENSG00000006071.7      E054    14.508353 0.01849662 17.59588 2.731798e-05
ENSG00000006625.13:E027 ENSG00000006625.13      E027     3.387866 0.23556958 17.21820 3.332288e-05
ENSG00000006744.14:E006 ENSG00000006744.14      E006     1.906249 0.49098505 17.18358 3.393576e-05
ENSG00000010244.12:E059 ENSG00000010244.12      E059    10.860515 0.09847943 23.51376 1.240234e-06
ENSG00000011465.12:E046 ENSG00000011465.12      E046    10.115533 0.00991492 17.57031 2.768780e-05
                               padj DAB2IP_high   DAB2IP_low log2fold_DAB2IP_high_DAB2IP_low genomicData
ENSG00000003402.15:E102 0.041529898   0.9379022 6.466887e-01                       1.1593566            
ENSG00000006071.7:E054  0.043587315   1.1932887 8.332773e-01                       1.3294337            
ENSG00000006625.13:E027 0.045175810   0.8723202 3.162357e-01                       2.5906126            
ENSG00000006744.14:E006 0.045488976   0.6990015 6.938573e-06                      17.9337945            
ENSG00000010244.12:E059 0.007795511   1.2766317 7.781390e-01                       1.8406042            
ENSG00000011465.12:E046 0.043786361   1.1024762 8.659708e-01                       0.8781098            
                        countData.1 countData.2 countData.3 countData.4 countData.5 countData.6 countData.7
ENSG00000003402.15:E102           5           3           5           2           8           0           3
ENSG00000006071.7:E054            0           4           1           6           6           2           6
ENSG00000006625.13:E027           0           1           0           4           0           0           1
ENSG00000006744.14:E006           0           0           0           0           0           0           0
ENSG00000010244.12:E059           3           3           3           7          13          11           1
ENSG00000011465.12:E046           0           3          11           3          13           2           5
                        countData.8 countData.9 countData.10 countData.11 countData.12 countData.13
ENSG00000003402.15:E102           1           5            1            7           11           13
ENSG00000006071.7:E054            0           4            1            5            4           48
ENSG00000006625.13:E027           4           2            1            4            5            6
ENSG00000006744.14:E006           0           0            0            0            6            2
ENSG00000010244.12:E059           3          10            1            7           12           47
ENSG00000011465.12:E046           5          11            9           11           11           33
                        countData.14 countData.15 countData.16 countData.17 countData.18 countData.19
ENSG00000003402.15:E102            3           12            5            5           12           14
ENSG00000006071.7:E054            95           74            5           13           67            2
ENSG00000006625.13:E027            1           11            4           11            7            2
ENSG00000006744.14:E006            0            8            0            9           10            0
ENSG00000010244.12:E059           12           21           10           28           23           11
ENSG00000011465.12:E046            1            0           22            2            4           23
                        countData.20  transcripts
ENSG00000003402.15:E102            8 ENSG0000....
ENSG00000006071.7:E054            22 ENSG0000....
ENSG00000006625.13:E027            7 ENSG0000....
ENSG00000006744.14:E006            8 ENSG0000....
ENSG00000010244.12:E059           11 ENSG0000....
ENSG00000011465.12:E046           25 ENSG0000....
sessionInfo()
R version 4.1.0 (2021-05-18)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Red Hat Enterprise Linux 8.5 (Ootpa)

Matrix products: default
BLAS/LAPACK: /nas/longleaf/rhel8/apps/r/4.1.0/lib/libopenblas_haswellp-r0.3.5.so

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

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

other attached packages:
 [1] DEXSeq_1.40.0               RColorBrewer_1.1-3          AnnotationDbi_1.56.2       
 [4] DESeq2_1.34.0               SummarizedExperiment_1.24.0 GenomicRanges_1.46.1       
 [7] GenomeInfoDb_1.30.1         IRanges_2.28.0              S4Vectors_0.32.4           
[10] MatrixGenerics_1.6.0        matrixStats_0.62.0          Biobase_2.54.0             
[13] BiocGenerics_0.40.0         BiocParallel_1.28.3         forcats_0.5.1              
[16] stringr_1.4.0               dplyr_1.0.9                 purrr_0.3.4                
[19] readr_2.1.2                 tidyr_1.2.0                 tibble_3.1.8               
[22] ggplot2_3.3.6               tidyverse_1.3.1             rhdf5_2.38.0               

loaded via a namespace (and not attached):
 [1] bitops_1.0-7           fs_1.5.2               lubridate_1.8.0        bit64_4.0.5           
 [5] filelock_1.0.2         progress_1.2.2         httr_1.4.2             tools_4.1.0           
 [9] backports_1.4.1        utf8_1.2.2             R6_2.5.1               DBI_1.1.2             
[13] colorspace_2.0-3       rhdf5filters_1.6.0     withr_2.5.0            prettyunits_1.1.1     
[17] tidyselect_1.1.2       curl_4.3.2             bit_4.0.4              compiler_4.1.0        
[21] cli_3.3.0              rvest_1.0.2            xml2_1.3.3             DelayedArray_0.20.0   
[25] scales_1.2.0           genefilter_1.76.0      rappdirs_0.3.3         Rsamtools_2.10.0      
[29] digest_0.6.29          XVector_0.34.0         pkgconfig_2.0.3        dbplyr_2.1.1          
[33] fastmap_1.1.0          rlang_1.0.4            readxl_1.3.1           rstudioapi_0.13       
[37] RSQLite_2.2.10         generics_0.1.2         hwriter_1.3.2          jsonlite_1.8.0        
[41] RCurl_1.98-1.6         magrittr_2.0.2         GenomeInfoDbData_1.2.7 Matrix_1.4-0          
[45] Rcpp_1.0.8.3           munsell_0.5.0          Rhdf5lib_1.16.0        fansi_1.0.3           
[49] lifecycle_1.0.1        stringi_1.7.6          zlibbioc_1.40.0        BiocFileCache_2.2.1   
[53] grid_4.1.0             blob_1.2.2             parallel_4.1.0         crayon_1.5.1          
[57] lattice_0.20-45        Biostrings_2.62.0      haven_2.4.3            splines_4.1.0         
[61] annotate_1.72.0        hms_1.1.1              KEGGREST_1.34.0        locfit_1.5-9.5        
[65] pillar_1.7.0           biomaRt_2.50.3         geneplotter_1.72.0     reprex_2.0.1          
[69] XML_3.99-0.9           glue_1.6.2             modelr_0.1.8           vctrs_0.4.1           
[73] png_0.1-7              tzdb_0.2.0             cellranger_1.1.0       gtable_0.3.0          
[77] assertthat_0.2.1       cachem_1.0.6           xtable_1.8-4           broom_1.0.0           
[81] survival_3.2-13        memoise_2.0.1          statmod_1.4.36         ellipsis_0.3.2
DESeq2 glmGamPoi DEXSeq • 2.3k views
ADD COMMENT
0
Entering edit mode

Thanks for posting and providing the reproducible example Jeremy.

ADD REPLY
2
Entering edit mode
Alejandro Reyes ★ 1.9k
@alejandro-reyes-5124
Last seen 3 months ago
Novartis Institutes for BioMedical Rese…

Hi Jeremy,

Thanks so much for the reproducible example. I've implemented the option to use glmGamPoi in the development version of DEXSeq. Good to have this functionality there in any case.

I've also ran a quick speed test based on your example, and it seems that we are not gaining speed with glmGamPoi (see code chunk below). I think this makes sense though, as the scalability issues of DEXSeq come from the fact that a coefficient is fitted for each sample, not because of data sparsity.

The default DEXSeq model was thought with small sample sizes and it's still one of the frontrunners in terms of performance. When analysing large datasets and computational resources are limited, one might consider exploring alternatives such as satuRn, diffSplice or even non parametric approaches for differential usage analyses.

Best regards, Alejandro

 > library(DEXSeq)                                                                                                                                                                                   
 > dxd <- readRDS("DEXSeq_Bioc_rand1.rds")                                                                                                                                                           
 > dxd <- estimateSizeFactors(dxd)                                                                                                                                                                   
 > system.time( dxd_glmgampoi <- estimateDispersions( dxd[1:1000,], fitType="glmGamPoi", quiet=TRUE ) )                                                                                              
    user  system elapsed
  50.780   0.244  51.143
 > system.time( dxd_default <- estimateDispersions( dxd[1:1000,] ) )                                                                                                                                 
    user  system elapsed
  23.097   0.033  23.185
!> sessionInfo()                                                                                                                                                                                     
 R Under development (unstable) (2023-01-16 r83624)
 Platform: x86_64-pc-linux-gnu (64-bit)
 Running under: CentOS Linux 7 (Core)

 Matrix products: default
 BLAS/LAPACK: /da/dmp/cb/reyesre5/software/miniconda3/envs/r_development/lib/libopenblasp-r0.3.21.so;  LAPACK version 3.9.0

 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

 time zone: Europe/Zurich
 tzcode source: system (glibc)

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

 other attached packages:
  [1] DEXSeq_1.45.2               RColorBrewer_1.1-3
  [3] AnnotationDbi_1.61.0        DESeq2_1.39.6
  [5] SummarizedExperiment_1.29.1 GenomicRanges_1.51.4
  [7] GenomeInfoDb_1.35.15        IRanges_2.33.0
  [9] S4Vectors_0.37.3            MatrixGenerics_1.11.0
 [11] matrixStats_0.63.0          Biobase_2.59.0
 [13] BiocGenerics_0.45.0         BiocParallel_1.33.9
ADD COMMENT
1
Entering edit mode

Very interesting work! I agree with all points made by Alejandro: having the design specified with formula( ~ sample + exon + Group:exon) will inevitably continue hurting scalability. I am curious to see to what extent glmGamPoi could save the day, but in the end doing this for many samples or e.g. single-cell data should remain infeasible.

I am not here to advertise my own method, but since Alejandro was already mentioning satuRn, I gave it a quick try (see below). Unfortunately I had to do it in R4.2, as R4.3 is currently giving me some issue. But anyway, running satuRn here takes under two seconds without parallellization, and should scale linearly w.r.t. the number of cells in the data. To give you an idea, I have performed an analysis on single-cell data in the past with approximately 2700 cells and say 20K features, which took me like ±5min without parallellization on my macbook pro laptop from 2019.

> library(DEXSeq)
> library(satuRn)
> dxd <- readRDS("DEXSeq_Bioc_rand1.rds")
> exonInfo <- rowData(dxd)
> colnames(exonInfo)[1:2] <- c("isoform_id", "gene_id")
> exonInfo$isoform_id <- rownames(exonInfo)
> sumExp <- SummarizedExperiment(assays = list(counts=featureCounts(dxd)),
+                                colData = sampleAnnotation(dxd),
+                                rowData = exonInfo)

> dxd <- estimateSizeFactors(dxd)

> system.time(dxd_default <- estimateDispersions(dxd[1:1000,]))
   user  system elapsed 
 43.173   0.237  43.639 

> system.time(satuRn::fitDTU(object = sumExp[1:1000,],
+                            formula = ~Group,
+                            verbose=FALSE))
   user  system elapsed 
  1.385   0.020   1.406
> sessionInfo()
R version 4.2.0 alpha (2022-04-04 r82084)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Big Sur 11.6

Matrix products: default
LAPACK: /Library/Frameworks/R.framework/Versions/4.2/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

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

other attached packages:
 [1] satuRn_1.7.2                DEXSeq_1.44.0               RColorBrewer_1.1-3          AnnotationDbi_1.60.0        DESeq2_1.38.3              
 [6] SummarizedExperiment_1.28.0 GenomicRanges_1.50.2        GenomeInfoDb_1.34.9         IRanges_2.32.0              S4Vectors_0.36.1           
[11] MatrixGenerics_1.10.0       matrixStats_0.63.0          Biobase_2.58.0              BiocGenerics_0.44.0         BiocParallel_1.32.5        

loaded via a namespace (and not attached):
 [1] bitops_1.0-7           bit64_4.0.5            filelock_1.0.2         progress_1.2.2         httr_1.4.5             rprojroot_2.0.3       
 [7] tools_4.2.0            utf8_1.2.3             R6_2.5.1               DBI_1.1.3              colorspace_2.1-0       tidyselect_1.2.0      
[13] prettyunits_1.1.1      bit_4.0.5              curl_5.0.0             compiler_4.2.0         cli_3.6.0              xml2_1.3.3            
[19] DelayedArray_0.24.0    scales_1.2.1           genefilter_1.80.3      pbapply_1.7-0          rappdirs_0.3.3         stringr_1.5.0         
[25] digest_0.6.31          Rsamtools_2.14.0       rmarkdown_2.20         XVector_0.38.0         htmltools_0.5.4        pkgconfig_2.0.3       
[31] dbplyr_2.3.0           fastmap_1.1.0          limma_3.54.1           rlang_1.0.6            rstudioapi_0.14        RSQLite_2.3.0         
[37] generics_0.1.3         hwriter_1.3.2.1        dplyr_1.1.0            RCurl_1.98-1.10        magrittr_2.0.3         GenomeInfoDbData_1.2.9
[43] Matrix_1.5-3           Rcpp_1.0.10            munsell_0.5.0          fansi_1.0.4            lifecycle_1.0.3        yaml_2.3.7            
[49] stringi_1.7.12         zlibbioc_1.44.0        BiocFileCache_2.6.1    grid_4.2.0             blob_1.2.3             parallel_4.2.0        
[55] crayon_1.5.2           lattice_0.20-45        Biostrings_2.66.0      splines_4.2.0          annotate_1.76.0        hms_1.1.2             
[61] KEGGREST_1.38.0        locfit_1.5-9.7         knitr_1.42             pillar_1.8.1           boot_1.3-28.1          geneplotter_1.76.0    
[67] codetools_0.2-19       biomaRt_2.54.0         XML_3.99-0.13          glue_1.6.2             evaluate_0.20          png_0.1-8             
[73] vctrs_0.5.2            locfdr_1.1-8           gtable_0.3.1           assertthat_0.2.1       cachem_1.0.7           ggplot2_3.4.1         
[79] xfun_0.37              xtable_1.8-4           survival_3.5-3         tibble_3.1.8           memoise_2.0.1          statmod_1.5.0         
[85] ellipsis_0.3.2
ADD REPLY
1
Entering edit mode

Thanks Jeroen (I recruited him to demo what speed improvement with satuRn would look like). Impressive.

ADD REPLY
0
Entering edit mode

Awesome, thanks for implementing Alejandro.

ADD REPLY
1
Entering edit mode

Thanks Jeroen and Alejandro! This is all very helpful.

ADD REPLY
2
Entering edit mode
@1625b20f
Last seen 13 months ago
Belgium

Hey Jeremy,

An alternative to estimating sample-specific coefficients, is the use of the (logarithm of the) total gene counts as offset in DESeq2/edgeR. This leads to an interpretation in terms of proportions (exon counts / gene counts), without estimating sample-specific coefficients and therefore scaling well with increasing sample size. To avoid division by 0, we add a pseudo count to every count and offset, of which the offset is equal to 0.

This was recently benchmarked by Jeroen Gilis and myself in our preprint (https://www.biorxiv.org/content/10.1101/2023.06.29.547014v1), achieving similar power compared to DEXSeq and satuRn on bulk RNA-seq, while being much faster. Note, however, that we noticed that satuRn still performed slightly better on scRNA-seq data. Possibly due to the issues with the large numbers of zeros in contemporary scRNA-seq platforms. Best regards, Alexandre

> library(DESeq2)
> library(DEXSeq)
> library(edgeR)
> library(dplyr)

> dxd <- readRDS("DEXSeq_Bioc_rand1")
> exonInfo <- rowData(dxd)
> colnames(exonInfo)[1:2] <- c("exon_id", "gene_id")
> exonInfo$exon_id <- rownames(exonInfo)
> dxd <- estimateSizeFactors(dxd)

> ddscounts <- featureCounts(dxd)
> ddsoffsetsgene <- aggregate(ddscounts,
+                        by = list("locus" = exonInfo$gene_id),
+                        FUN = sum) 

> ddsoffsets <- ddsoffsetsgene[match(exonInfo$gene_id,ddsoffsetsgene$locus),] %>% mutate("locus" = NULL)
> ddscounts[ddsoffsets == 0] <- 1
> ddsoffsets[ddsoffsets == 0] <- 1

> dds <- DESeqDataSetFromMatrix(countData = ddscounts,
+                                     colData = sampleAnnotation(dxd),
+                                     rowData = exonInfo,
+                                     design= ~ Group)

> normalizationFactors(dds) <- ddsoffsets %>% as.matrix()

> DGE <- DGEList(counts=ddscounts)
> DGE$offset <- log(ddsoffsets %>% as.matrix())
> design <- model.matrix(~Group, data = colData(dxd) %>% 
+                            as.data.frame() %>% 
+                            filter(exon == "this"))


> system.time(dxd <- DEXSeq::estimateDispersions(dxd[1:1000,]))
   user  system elapsed 
 34.976   0.447  35.714 
> system.time(dds <- DESeq2::estimateDispersionsGeneEst(dds[1:1000,]))
found already estimated gene-wise dispersions, removing these
   user  system elapsed 
  0.400   0.010   0.416 
> system.time(DGE <- edgeR::estimateDisp(DGE[1:1000,], design = design))
   user  system elapsed 
  0.097   0.003   0.100 


> sessionInfo()
R version 4.1.0 (2021-05-18)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS 12.5.1

Matrix products: default
LAPACK: /Library/Frameworks/R.framework/Versions/4.1/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

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

other attached packages:
 [1] dplyr_1.1.2                 edgeR_3.36.0                limma_3.50.3                DEXSeq_1.40.0              
 [5] RColorBrewer_1.1-3          AnnotationDbi_1.56.2        BiocParallel_1.28.3         DESeq2_1.34.0              
 [9] SummarizedExperiment_1.24.0 Biobase_2.54.0              MatrixGenerics_1.6.0        matrixStats_0.63.0         
[13] GenomicRanges_1.46.1        GenomeInfoDb_1.30.1         IRanges_2.28.0              S4Vectors_0.32.4           
[17] BiocGenerics_0.40.0        

loaded via a namespace (and not attached):
 [1] httr_1.4.6             bit64_4.0.5            splines_4.1.0          statmod_1.5.0          BiocFileCache_2.2.1   
 [6] blob_1.2.4             Rsamtools_2.10.0       GenomeInfoDbData_1.2.7 progress_1.2.2         yaml_2.3.7            
[11] pillar_1.9.0           RSQLite_2.3.1          lattice_0.21-8         glue_1.6.2             digest_0.6.31         
[16] XVector_0.34.0         colorspace_2.1-0       htmltools_0.5.5        Matrix_1.5-1           XML_3.99-0.14         
[21] pkgconfig_2.0.3        biomaRt_2.50.3         genefilter_1.76.0      zlibbioc_1.40.0        xtable_1.8-4          
[26] scales_1.2.1           tibble_3.2.1           annotate_1.72.0        KEGGREST_1.34.0        generics_0.1.3        
[31] ggplot2_3.4.2          cachem_1.0.8           cli_3.6.1              survival_3.5-5         magrittr_2.0.3        
[36] crayon_1.5.2           memoise_2.0.1          evaluate_0.21          fansi_1.0.4            xml2_1.3.4            
[41] hwriter_1.3.2.1        prettyunits_1.1.1      tools_4.1.0            hms_1.1.3              lifecycle_1.0.3       
[46] stringr_1.5.0          munsell_0.5.0          locfit_1.5-9.7         DelayedArray_0.20.0    Biostrings_2.62.0     
[51] compiler_4.1.0         rlang_1.1.1            grid_4.1.0             RCurl_1.98-1.12        rstudioapi_0.14       
[56] rappdirs_0.3.3         bitops_1.0-7           rmarkdown_2.21         gtable_0.3.3           curl_5.0.0            
[61] DBI_1.1.3              R6_2.5.1               knitr_1.43             fastmap_1.1.1          bit_4.0.5             
[66] utf8_1.2.3             filelock_1.0.2         stringi_1.7.12         parallel_4.1.0         Rcpp_1.0.10           
[71] vctrs_0.6.2            geneplotter_1.72.0     png_0.1-8              dbplyr_2.3.2           tidyselect_1.2.0      
[76] xfun_0.39
ADD COMMENT

Login before adding your answer.

Traffic: 350 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6