scrublet in seurat pipeline

snippet_scrublet_seurat.Rmd

```{r}
library(reticulate)
# You can set the conda environment
use_condaenv("scrublet", required = T)
```

```{python}
import scrublet as scr
import scipy.io
import matplotlib.pyplot as plt
import numpy as np
import os

def predict_doublets(matrix):
    counts_matrix = scipy.io.mmread(matrix).T.tocsc()
    scrub = scr.Scrublet(counts_matrix, expected_doublet_rate=0.10)
    doublet_scores, predicted_doublets = scrub.scrub_doublets(min_counts=2, min_cells=3, min_gene_variability_pctl=85, n_prin_comps=30)
    #scrub.plot_histogram()
    #scrub.set_embedding('UMAP', scr.get_umap(scrub.manifold_obs_, 10, min_dist=0.3))
    #scrub.plot_embedding('UMAP', order_points=True)
    return scrub.doublet_scores_obs_
    

matrix = "../data/00_filtered_feature_bc_matrix/count_Mutant/filtered_feature_bc_matrix/matrix.mtx"
ds_mt = predict_doublets(matrix)

matrix = "../data/00_filtered_feature_bc_matrix/count_Wild-type/filtered_feature_bc_matrix/matrix.mtx"
ds_wt = predict_doublets(matrix)
```

```{r}
ds = c(py$ds_mt, py$ds_wt)
names(ds) = rownames(so[["orig.ident"]]) 
# so is a seurat object
so[['doublet_score']] = ds
```