DESeqDatasetFromMatrix() doesn’t like tibbles?

[cansavvy]

There were 3 or 4 participants who encountered this problem at March 2021 workshops.

They had a tibble and tried to turn it into a DESeqDataSet. The error returned is something about row.names() but an as.data.frame() solves the problem. Idk that this requires a change in instruction but it was a common problem so I figured we should have it written down.

[jashapiro]

I would lean toward changing to a matrix, since that is what the function explicitly asks for. This is also what we model in the exercise notebook:

training-modules/RNA-seq/06-openpbta_heatmap.Rmd

Lines 127 to 155 in fa3abe5

	```{r convert_round, live = TRUE}
	rnaseq_mat <- rnaseq_exp %>%
	# move gene_id to the rownames
	tibble::column_to_rownames("gene_id") %>%
	# convert to a matrix and round
	as.matrix() %>%
	round()
	```
	### Variance Stabilizing Transformation
	Raw counts are not usually suitable for the algorithms we use for clustering and heatmap display, so we will use the `vst()` function from the `DESeq2` package to transform our data.
	Since we are starting from a matrix, not a `SummarizedExperiment` as we did previously, we will need to provide the sample information ourselves.
	Just to be sure nothing is out of order, we will check that the identifiers for the sample information stored in `histologies_df` matches the columns of our matrix.
	```{r check-order}
	all.equal(histologies_df$Kids_First_Biospecimen_ID,
	colnames(rnaseq_mat))
	```
	Now we can make our matrix into a `DESeq2` dataset, adding on the sample information from `histologies_df`.
	Unlike when we were performing differential expression analysis, we won't provide an experimental design at this stage.
	```{r make-DESEq}
	ddset <- DESeqDataSetFromMatrix(rnaseq_mat,
	colData = histologies_df,
	design = ~ 1) # don't store an experimental design
	```