Hi,
Not sure if anyone have similar result. When I look into my output fcs file, I just realized that the XGBoost marker are not even correlated with it's corresponding marker in that sample. Thereotically, the expression of this XGBoost marker are imputated via cells in this sample, so the imputated value must be highly correlated with it's ground-truth value. But in my result, the imputated value are entirely negtive and not correlated with ground-truth value at all.
Any suggestions?