Hello All,

I have generated a count matrix using kallisto bustools, and i would like to remove empty droplets based on knee plot.I would like to know how to identify i) Inflection ii) Knee point threshold cutoff using python code.

i know we can use "dropletutils" from R., but i would like to see any module available to identify the knee cut-off for sparse expression matrix using python. Please help me!!!

enter code here
 import numpy as np
 import pandas as pd
 import scanpy as sc
 import anndata
 from sklearn.decomposition import TruncatedSVD
 import matplotlib
 import matplotlib.pyplot as plt
  import scrublet as scr
  from scipy import sparse, io
  import statistics
   list1=[f1,f2,f3,f4,f5]
   for i in range(len(list1)):
   #read Anndata
        adata= anndata.read_h5ad(~/Akila/unsplice_counts/"+str(list1[i])+".h5ad")
   # # Create the flipped and rotated "knee plot"
        knee = np.sort((np.array(adata.X.sum(axis=1))).flatten())[::-1]
        fig, ax = plt.subplots(figsize=(10, 7))
        ax.loglog(range(len(knee)), knee,linewidth=5, color="g")
        ax.set_xlabel("Set of Barcodes")
        ax.set_ylabel("UMI Counts")
        plt.grid(True, which="both")
# calculate stats on knee 
        print(list1[i],min(knee),np.median(knee),max(knee),np.mean(knee))
#save plot
       plt.savefig("~/Akila/unsplice_knee/"+str(list1[i])+".png")

Thanks Akila