Unassigned cells

[hxy265]

Hi thanks for making a very cool package!

I have 10X scDNA-seq data with matched 10X scRNA-seq cells. I'm running Treealign for total CN with scDNA-seq clones already labelled. It's performing reasonably well although i have around 20 subclones so it is quite challenging.

However, I'm getting around 30% of scRNA-seq as unassigned. Visually the inferCNV tracks of these cells look similar to cells that are successfully assigned. So i'm wondering why these are unassigned? Are there some parameters i should be changing? Thanks!

[adamcweiner]

Sorry for not responding earlier but yes there are parameters you can tune if you believe the model is leaving too many scRNA cells as unassigned under default settings. For a single RNA cell, if the clone assignment probability is <0.8 or the clone assignments are not consistent in >70% of repeated runs, the cell will be denoted as unassigned. These two thresholds correspond to the min_clone_assign_prob and min_clone_assign_freq arguments of the CloneAlign and CloneAlignTree classes which can be tuned by the user. For both parameters, setting lower values will produce fewer unassigned cells.

TreeAlign/src/treealign/clonealign.py

Line 498 in e8d011d

current_clone_assign_discrete[current_clone_assign_prob < self.min_clone_assign_prob] = np.nan

TreeAlign/src/treealign/clonealign.py

Line 535 in e8d011d

clone_assign_max[clone_assign_freq < self.min_clone_assign_freq] = np.nan