directionality of log2FoldChange (DESeq2) opposite to voom and edgeR about tidybulk HOT 11 CLOSED

Thanks for the coding note. I'm still not used to the tidy way, so much appreciated.

from tidybulk.

That worked just fine, excellent :) Nice work.

I would still change the error message to be:

tidybulk says: for DESeq2 the list of constrasts should be given in the form list(c("condition_column","condition1","condition2")) i.e. list(c("genotype","knockout","wildtype")).

rather than

tidybulk says: for DESeq2 the list of constrasts should be given of the form list(c("condition","C","B"))

from tidybulk.

Hello @AMChalkie . We though we fixed it. Indeed the workshop material shows expected results.

Could you please execute the section

https://stemangiola.github.io/biocasia2020_tidytranscriptomics/articles/tidytranscriptomics.html#differential-expression-1

and tell us if you still see the opposite results. In this way we can understand where the issue lies.

[EDIT]

I see, you are defining contrasts. Sorry this is bad bug! I will take care of it ASAP, also adding those tests.

from tidybulk.

Some notes on coding.

You don't necessarily need to replicate code. This would suffice:

counts_de.voom <-
   tt.aggr %>% 

   identify_abundant(factor_of_interest = genotype) %>% 

   test_differential_abundance(
      .formula = ~ 0 + genotype, # + cell,
      .contrasts = c("genotypeKI - genotypeWT"),
      method = "limma_voom", # edgeR_quasi_likelihood", # Alternative method. "limma_voom",
      omit_contrast_in_colnames = TRUE,
      prefix= "limma_voom."
   ) %>%

   test_differential_abundance(
      .formula = ~ 0 + genotype, # + cell,
.      contrasts = c("genotypeKI - genotypeWT"),
      method = "edgeR_quasi_likelihood", # Alternative method. "limma_voom",
      omit_contrast_in_colnames = TRUE,
      prefix= "edgeR_quasi_likelihood."
  ) %>%

  test_differential_abundance(
    .formula = ~ 0 + genotype, # + cell,
.    contrasts = c("genotypeKI - genotypeWT"),
    method = "DESeq2", # Alternative method. "limma_voom",
    omit_contrast_in_colnames = TRUE,
    prefix= "DESeq2."
 )

from tidybulk.

Hello @AMChalkie I have created a pull request. Could you please check that the respective branch leads to expected results on your side?

That would be much appreciated.

from tidybulk.

Not yet:

tidybulk says: All methods use raw counts,
irrespective of if scale_abundance or adjust_abundance have been calculated,
therefore it is essential to add covariates such as batch effects (if applicable) in the formula.
Error in get_differential_transcript_abundance_deseq2(., .formula, .sample = !!.sample, :
tidybulk says: for the moment, the .contrasts argument for
DESeq2 gets directly passed to DESeq2 method.
Contrasts must be a list that contains contrasts of
the form c("condition","C","B")
Called from: get_differential_transcript_abundance_deseq2(., .formula, .sample = !!.sample,
.transcript = !!.transcript, .abundance = !!.abundance, .contrasts = .contrasts,
method = method, scaling_method = scaling_method, omit_contrast_in_colnames = omit_contrast_in_colnames,
prefix = prefix)

from tidybulk.

Maybe I should improve this message

for the moment, the .contrasts argument for
DESeq2 gets directly passed to DESeq2 method.
Contrasts must be a list that contains contrasts of
the form c("condition","C","B")

Is not self-explanatory? Feel also free to edit it directly in the repo if you can think about something clearer for users.

from tidybulk.

For documentation:

for the moment, the .contrasts argument for
DESeq2 gets directly passed to DESeq2 method.
Contrasts must be a list that contains contrasts of
the form c("factorName","numeratorLevel","denominatorLevel")

I would have expected that this would work, but no...:

tt.aggr %>% identify_abundant(factor_of_interest = genotype) %>%
test_differential_abundance(
.formula = ~ 0 + genotype,
.contrasts = c('genotype','KI','WT'),
method = "DESeq2",
omit_contrast_in_colnames = TRUE,
prefix= "DESeq2."
)
tidybulk says: All methods use raw counts,
irrespective of if scale_abundance or adjust_abundance have been calculated,
therefore it is essential to add covariates such as batch effects (if applicable) in the formula.
estimating size factors
estimating dispersions
gene-wise dispersion estimates
mean-dispersion relationship
final dispersion estimates
fitting model and testing
Error in checkContrast(contrast, resNames) :
'contrast', as a character vector of length 3, should have the form:
contrast = c('factorName','numeratorLevel','denominatorLevel'),
see the manual page of ?results for more information
In addition: Warning message:
In get_differential_transcript_abundance_deseq2(., .formula, .sample = !!.sample, :
tidybulk says: you can omit contrasts in column names only when maximum one contrast is present
Called from: checkContrast(contrast, resNames)

Also, this is after removing the check for equals("list") as it failed here otherwise.

.contrasts %>% is.null %>% not() &
.contrasts %>% class %>% equals("list") #%>% not()

from tidybulk.

"a list that contains contrasts"

list(c('genotype','KI','WT'))

As you can specify many contrast combinations (similarly to edgeR)

Maybe I should be more explicit

from tidybulk.

I see the other problem. Tomorrow I fix everything

from tidybulk.

Hello @AMChalkie could you please reinstall this branch and try again? I have improve error messages. Let's see if it guides you better.

from tidybulk.