dviraran
commented
4 years ago Hi,
Sorry for the late response. Missed this issue. Anyway, this repository is not supported any more. Please use https://github.com/LTLA/SingleR to report issues.
Best, Dvir
Hi,
code in fineTuningRound method:
ref_data.filtered = ref_data.filtered[genes.filtered,] sc_data.filtered = as.matrix(sc_data[genes.filtered]) if (length(genes.filtered)<20) { return (topLabels[1]) } if (sd(sc_data.filtered)>0) { r=cor.stable(sc_data.filtered,ref_data.filtered,method='spearman') agg_scores = quantileMatrix(r,types.filtered,quantile.use)[1,]; agg_scores = sort(agg_scores,decreasing = T) agg_scores = agg_scores[-length(agg_scores)] topLabels = names(agg_scores)[agg_scores>=agg_scores[1]-fine.tune.thres] }
sc_data.filtered = as.matrix(sc_data[genes.filtered]) missed a ',' , which should be sc_data.filtered = as.matrix(sc_data[genes.filtered, ]) ,so actually the second round with top labels was not run.
what's more, agg_scores = quantileMatrix(r,types.filtered,quantile.use)[1,] didn't get the right cell that needs to 'FineTune ', here recalculate the correlation of sc_data.filtered and ref_data.filtered which could be time cosuming, just need to pick up the cell in sc_data.filtered to do the correlation
Bing