scran "sizes" parameter

[drisso]

Currently, the scran wrapper in SCONE uses a value of sizes that is not the default in scran:
https://github.com/YosefLab/scone/blob/master/R/SCONE_DEFAULTS.R#L136

@mbcole what is the rationale for this value? Why not using the scran default seq(20, 100, 5)?

[drisso]

scran_fn <- function(ei) {
  sce <- newSCESet(countData=data.frame(ei))
  sce <- computeSumFactors(sce, sizes = seq(20, NCOL(ei), 5))
  sce <- normalize(sce)
  eo <- norm_exprs(sce)
  return(2^eo - 1)
}

scran_cluster_fn <- function(ei) {
  sce <- newSCESet(countData=data.frame(ei))
  cl <- quickCluster(sce, min.size = 50)
  sce <- computeSumFactors(sce, clusters = cl, sizes = min(table(cl)))
  sce <- normalize(sce)
  eo <- norm_exprs(sce)
  return(2^eo - 1)
}

[mbcole]

@drisso what's the motivation behind seq(20, NCOL(ei), 5) in scran_fn?

Why not use the default you cited above?

[drisso]

I was using the default, which is sizes=seq(20, 100, 5), but I've changed it for the fluidigm data that has only ~70 cells. I guess a better way would be sizes=seq(20, min(100, NCOL(ei)), 5)...

[ghost]

I am trying to learn how to use scran with scone. What is newSCESet here?