Issue with evaluate_uncertainty(), no difference in sample proportion mean information

[audrey-truong]

Hey all! I am having trouble/question about the evaluate_uncertainty() function. As I run my data through the infer_tree_structure() while letting the algorithm deciding the number of cluster for me, the res object from that output were used to put in evaluate_uncertainty(). The output for that(the sample proportion mean information) indicates that one branch has all the samples. My code and output for that is here:
CODE

infer tree structure based on origin cell type

set.seed(12345)
d27_res_cell_type = infer_tree_structure(pca = d27_lamian_data[['pca']],
expression = d27_lamian_data[['expression']],
cellanno = d27_lamian_data[['cellanno']],
origin.celltype = c('Neural Progenitor cells'),
#number.cluster = 10,
xlab='Principal component 1',
ylab = 'Principal component 2')
result_cell_type <- evaluate_uncertainty(d27_res_cell_type, n.permute=100)
result_cell_type[[2]]

OUPUT

result_cell_type[[2]]
27_Enriched_7 27_Enriched_16 27_Enriched_11 27_Enriched_19 27_Enriched_3
c(4, 8, 1) 1 1 1 1 1
c(4, 2, 7, 6, 3, 5) 0 0 0 0 0
27_Enriched_9 27_Enriched_17 27_Enriched_8 27_Enriched_15 27_Enriched_0
c(4, 8, 1) 1 1 1 1 1
c(4, 2, 7, 6, 3, 5) 0 0 0 0 0
27_Enriched_13 27_Nonenriched_6 27_Nonenriched_2 27_Nonenriched_16
c(4, 8, 1) 1 1 1 1
c(4, 2, 7, 6, 3, 5) 0 0 0 0
27_Nonenriched_10 27_Nonenriched_8 27_Nonenriched_14 27_Nonenriched_7
c(4, 8, 1) 1 1 1 1
c(4, 2, 7, 6, 3, 5) 0 0 0 0
27_Nonenriched_19 27_Nonenriched_4
c(4, 8, 1) 1 1
c(4, 2, 7, 6, 3, 5)

This shows that the branch 4,8,1 have all the proportions. This causes module 2 to fail. However, as I try to infer_tree_structure while specifying the number of clusters(which I got from the clustering from the Seurat pipeline which is 10), the proportion showed some differences:
CODE
d27_res_cell_type1 = infer_tree_structure(pca = d27_lamian_data[['pca']],
expression = d27_lamian_data[['expression']],
cellanno = d27_lamian_data[['cellanno']],
origin.celltype = c('Neural Progenitor cells'),
number.cluster = 10,
xlab='Principal component 1',
ylab = 'Principal component 2')
result_cell_type1 <- evaluate_uncertainty(d27_res_cell_type1, n.permute=100)
result_cell_type1[[2]]

OUPUT

result_cell_type1[[2]]
27_Enriched_7 27_Enriched_16 27_Enriched_11 27_Enriched_19 27_Enriched_3
c(4, 8, 1) 0.2304833 0.2466488 0.17325228 0.19196429 0.2222222
4:5 0.2156134 0.1568365 0.09422492 0.18526786 0.1522634
c(5, 10, 6, 3, 7, 2) 0.4783147 0.5348525 0.68085106 0.55803571 0.5637860
c(5,9) 0.0755886 0.0616622 0.05167173 0.06473214 0.0617284
27_Enriched_9 27_Enriched_17 27_Enriched_8 27_Enriched_15 27_Enriched_0
c(4, 8, 1) 0.22551929 0.21296296 0.1596154 0.25303644 0.21052632
4:5 0.14540059 0.16481481 0.1826923 0.15587045 0.13716108
c(5, 10, 6, 3, 7, 2) 0.56083086 0.57037037 0.5826923 0.55060729 0.57416268
c(5,9) 0.06824926 0.05185185 0.0750000 0.04048583 0.07814992
27_Enriched_13 27_Nonenriched_6 27_Nonenriched_2 27_Nonenriched_16
c(4, 8, 1) 0.22745098 0.16503122 0.1897019 0.13953488
4:5 0.16274510 0.10347904 0.1481481 0.11849391
c(5, 10, 6, 3, 7, 2) 0.53921569 0.67082962 0.5790425 0.64230343
c(5,9) 0.07058824 0.06066012 0.0831075 0.09966777
27_Nonenriched_10 27_Nonenriched_8 27_Nonenriched_14 27_Nonenriched_7
c(4, 8, 1) 0.18757613 0.1414474 0.13725490 0.10227273
4:5 0.10962241 0.1101974 0.11764706 0.09318182
c(5, 10, 6, 3, 7, 2) 0.64677223 0.6858553 0.66078431 0.73409091
c(5,9) 0.05602923 0.0625000 0.08431373 0.07045455
27_Nonenriched_19 27_Nonenriched_4
c(4, 8, 1) 0.17556468 0.17986799
4:5 0.12217659 0.12541254
c(5, 10, 6, 3, 7, 2) 0.62114990 0.63036304
c(5,9) 0.08110883 0.06435644

Does anyone have any idea for why this is happening or any recommendations of how to proceed?