CreateChromatinAssay min.cells and min.features behavior?

[varemo]

Hi,

Thanks for a great package!

I have a question about the min.cells and min.featuresarguments.

Let's say I have a counts matrix (feature-barcode matrix) in R with 115,468 features (rows) and 176,355 cells (columns). Let's further assume that 14 features have no counts in any cells and that 96,233 cells have zero counts. See below:

> dim(counts)
[1] 115468 176355

> table(rowSums(counts)==0)

 FALSE   TRUE 
115454     14

> table(colSums(counts)==0)

FALSE  TRUE 
80122 96233 

Running CreateChromatinAssay with default values for min.features and min.cells return an object with 115,468 features and 80,122 cells. That is, all features (including the 14 with no counts) are kept, whereas the 96,233 zero-count cells are removed (see below). The docs for min.features states "Include cells where at least this many features are detected." At least zero, would imply inclusion of cells with zero counts, at least the way I interpret this.

> CreateChromatinAssay(counts, fragments = frags, min.features = 0, min.cells = 0)
ChromatinAssay data with 115468 features for 80122 cells
Variable features: 0 
Genome: 
Annotation present: FALSE 
Motifs present: FALSE 
Fragment files: 1 

Reading the docs I would assume that setting min.features=1 would result in including cells with at least 1 detected feature. However, with this setting only 67,155 cells are kept. These cells, in fact, have >1 counts.

> CreateChromatinAssay(counts, fragments = frags, min.features = 1, min.cells = 0)
ChromatinAssay data with 115468 features for 67155 cells
Variable features: 0 
Genome: 
Annotation present: FALSE 
Motifs present: FALSE 
Fragment files: 1 

> table(colSums(counts)>1)

 FALSE   TRUE 
109200  67155 

The min.cells argument works as intended, at least in the way I interpret it, as setting it to 1 keeps all features with at least 1 count (i.e. removes the 14 features with zero counts and keeping the remaining 115,454):

> CreateChromatinAssay(counts, fragments = frags, min.features = 1, min.cells = 1)
ChromatinAssay data with 115454 features for 67155 cells
Variable features: 0 
Genome: 
Annotation present: FALSE 
Motifs present: FALSE 
Fragment files: 1 

Is this behavior intended? (Maybe I am missing the definition of "detected"?)

To me it seems that the behavior described in the docs should be:

counts <- counts[rowSums(counts) >= min.cells, colSums(counts) >= min.features]

but currently is:

counts <- counts[rowSums(counts) >= min.cells, colSums(counts) > min.features]

Anyone has any comments on this?

[timoast]

You're right, it should be >= min.features in the code. I've updated this on the develop branch.

[varemo]

Thanks Tim!