Error in bcRanks

[mj-jimenez]

bcRanks returns the following error:

Error in quantile.default(as.numeric(out$residuals.mean), prob = seq(from = 0,  : 
  missing values and NaN's not allowed if 'na.rm' is FALSE

I think it is due to bad cell subsetting.

[mj-jimenez]

Actually, it is due to all cells in a cluster having NaN values for a signature. To solve this there are several approaches:

• Imputation of missing values after bcScore. Right now, it has to be done as part of regression. We should separate these two functions.
• Convert NaNs to 0 within bcRanks (to compute the residuals).
• Probably the best one: Assign NaN to 0s (related to #113).

[SGMartin]

Any news on this front @mj-jimenez ?

[mj-jimenez]

This error appears in bcRanks but it is caused by the NaN values introduced when creating the beyondcell object. As said in #113, the best practice right now is this work-around until we update the bcScore function:

# Beyondcell object with a lot of NaN values
bcobj <- bcScore(sc, SSc, expr.thres = 0.1)

# Filter out spots with a high percentage of NAs                                                                                                                                                                                       
bcobj.filtered <- bcSubset(bcobj, nan.cells = 0.95)                                                                                                                                                                                    

# Replace NAs by 0s                                                                                                                                                                                                                    
bcobj.filtered@normalized[is.na(bcobj.filtered@normalized)] <- 0                                                                                                                                                                       
bcobj.recomputed <- bcRecompute(bcobj.filtered, slot = "normalized") 

[mj-jimenez]

Also, this problem is signature related (i.e. most of the genes in the signatures are not found in the expression matrix). This happened in beyondcell v2.0, which contained a high proportion of non-coding genes. Since beyondcell v2.1, the signatures only contain coding genes, thus this error is unlikely to happen with a standard single-cell experiment.