Select Markers¶
This tutorial shows how to select markers with the scmags package.
Let’s perform these operations with the baron_h1 data set in the package.
For this, we first import the package.
[1]:
import scmags as mg
Then we can start the operations by loading the dataset.
[2]:
baron_h1 = mg.datasets.baron_h1()
Filter Genes¶
First, redundant genes need to be filtered out for computational efficiency.
[3]:
baron_h1.filter_genes()
-> Eliminating low expression genes
-> Selecting cluster-specific candidate marker genes
Here’s how you can view the remaining genes after filtering
At the output of this function, a dictionary structure is returned, and each key corresponds to a cluster.
[4]:
rem_genes = baron_h1.get_filter_genes()
dict(list(rem_genes.items())[0:2])
[4]:
{'acinar': ['ALB',
'ALDOB',
'CEL',
'CELA2A',
'CUZD1',
'GP2',
'KLK1',
'PDIA2',
'PNLIPRP1',
'PNLIPRP2'],
'activated_stellate': ['ADAMTS12',
'COL6A3',
'CRLF1',
'FBN1',
'FMOD',
'GGT5',
'LAMC3',
'SFRP2',
'THBS2',
'VCAN']}
If you want, you can display the corresponding indexes in the data matrix.
[5]:
baron_h1.get_filter_genes(ind_return=True)
[5]:
{'acinar': array([ 550, 572, 3055, 3057, 4062, 6872, 8968, 12577, 13082,
13083]),
'activated_stellate': array([ 276, 3637, 3836, 5878, 6179, 6624, 9231, 15293, 17233,
18923]),
'alpha': array([ 2390, 6534, 7966, 8442, 8713, 10734, 11417, 12416, 14361,
15835]),
'beta': array([ 68, 322, 1974, 4675, 5342, 6383, 7240, 11265, 13322,
15686]),
'delta': array([ 1449, 4516, 6407, 6543, 9345, 10318, 12504, 14210, 16021,
16794]),
'ductal': array([ 3383, 4203, 9048, 9077, 10397, 13242, 13639, 15205, 16828,
17311]),
'endothelial': array([ 357, 3402, 4968, 5360, 6159, 8756, 12396, 14586, 16167,
19048]),
'epsilon': array([ 6027, 6299, 6358, 6637, 9723, 11320, 12939, 15274, 16388,
19024]),
'gamma': array([ 932, 6357, 9325, 10138, 11153, 12375, 13822, 16601, 18319,
18672]),
'macrophage': array([ 1971, 1973, 3888, 4127, 7307, 7541, 8490, 12935, 15095,
15487]),
'mast': array([ 63, 252, 2020, 2816, 3734, 6502, 8896, 14929, 17869,
17870]),
'quiescent_stellate': array([ 315, 372, 3930, 4987, 6668, 8266, 8633, 11057, 14362,
19198]),
'schwann': array([ 1728, 3859, 4579, 6356, 6911, 7961, 8585, 15146, 16160,
16509]),
't_cell': array([ 2729, 2842, 2860, 6503, 8385, 12389, 16339, 17887, 18058,
19329])}
If you have not set an intra-cluster expression rate threshold, you can also view the automatically determined thresholds.
[6]:
baron_h1.get_filt_cluster_thresholds
[6]:
{'acinar': 0.6545454561710358,
'activated_stellate': 0.7058823555707932,
'alpha': 0.6525423675775528,
'beta': 0.6416284441947937,
'delta': 0.6355140209197998,
'ductal': 0.6791666597127914,
'endothelial': 0.6192307695746422,
'epsilon': 0.6538461595773697,
'gamma': 0.6285714358091354,
'macrophage': 0.6785714328289032,
'mast': 0.625,
'quiescent_stellate': 0.614130437374115,
'schwann': 0.7000000029802322,
't_cell': 0.75}
If you don’t want automatic threshold determination for filtering, you can set a threshold yourself.
This value should be between 0-1.
[7]:
baron_h1.filter_genes(in_cls_thres=0.7)
-> Eliminating low expression genes
-> Selecting cluster-specific candidate marker genes
As can be seen, the threshold was set as 70% for all cells.
In this case, genes expressed in less than 70% of the cells in the cluster are filtered out.
[8]:
baron_h1.get_filt_cluster_thresholds
[8]:
{'acinar': 0.7,
'activated_stellate': 0.7,
'alpha': 0.7,
'beta': 0.7,
'delta': 0.7,
'ductal': 0.7,
'endothelial': 0.7,
'epsilon': 0.7,
'gamma': 0.7,
'macrophage': 0.7,
'mast': 0.7,
'quiescent_stellate': 0.7,
'schwann': 0.7,
't_cell': 0.7}
You can also set the number of genes that will remain after filtering.
This may be necessary when selecting more marker genes.Because marker selection is carried out on the remaining genes after filtering.
As can be seen in the example, when the parameter is set to 20, 20 genes remain after filtering for each cluster.
[9]:
baron_h1.filter_genes(nof_sel=20)
baron_h1.get_filter_genes(ind_return=True)
-> Eliminating low expression genes
-> Selecting cluster-specific candidate marker genes
[9]:
{'acinar': array([ 550, 572, 799, 3055, 3057, 3733, 4020, 4062, 6872,
7103, 7104, 8968, 10655, 10732, 12577, 13081, 13082, 13083,
13620, 16742]),
'activated_stellate': array([ 276, 405, 1990, 3632, 3633, 3637, 3836, 5878, 6179,
6182, 6624, 9231, 9504, 9691, 10387, 15293, 17233, 17850,
18923, 19025]),
'alpha': array([ 2390, 3235, 6534, 6730, 7966, 8442, 8713, 10734, 11417,
12291, 12339, 12416, 12508, 12637, 14361, 15275, 15835, 15991,
17366, 18944]),
'beta': array([ 68, 322, 1974, 2995, 3533, 4675, 4993, 5342, 6383,
6420, 7240, 10627, 11265, 12321, 13322, 13959, 14098, 14999,
15686, 18304]),
'delta': array([ 959, 1449, 1557, 2353, 4516, 6407, 6543, 9345, 10068,
10318, 11279, 12285, 12504, 12506, 14210, 15098, 16021, 16729,
16794, 17839]),
'ductal': array([ 458, 809, 3383, 3401, 4203, 4835, 5129, 6855, 8818,
9048, 9077, 9840, 10397, 13242, 13639, 15205, 15912, 16828,
17129, 17311]),
'endothelial': array([ 357, 2879, 3402, 4968, 5356, 5360, 6159, 7892, 8756,
12361, 12396, 12627, 13041, 13115, 14586, 14892, 16167, 16564,
17277, 19048]),
'epsilon': array([ 190, 878, 2383, 2979, 6027, 6212, 6299, 6358, 6637,
9723, 11148, 11320, 11473, 12939, 13229, 13895, 15211, 15274,
16388, 19024]),
'gamma': array([ 932, 2159, 3236, 6357, 7869, 9325, 10138, 11153, 11391,
12296, 12375, 12507, 12630, 13386, 13822, 14998, 16601, 17665,
18319, 18672]),
'macrophage': array([ 482, 1971, 1973, 2014, 3888, 4127, 7307, 7535, 7536,
7541, 7545, 7547, 8490, 11002, 12935, 15095, 15160, 15487,
15841, 16307]),
'mast': array([ 63, 252, 290, 1026, 2020, 2816, 3734, 6502, 7308,
8896, 9696, 13773, 14118, 14396, 14929, 15163, 15544, 15773,
17869, 17870]),
'quiescent_stellate': array([ 274, 287, 315, 372, 1713, 1823, 3930, 4987, 6668,
7906, 8247, 8266, 8469, 8633, 8922, 11057, 11455, 14362,
17186, 19198]),
'schwann': array([ 1728, 3859, 4579, 4988, 5199, 5551, 6356, 6911, 7961,
8585, 10122, 10469, 11171, 11492, 12598, 13240, 14221, 15146,
16160, 16509]),
't_cell': array([ 2729, 2840, 2842, 2854, 2860, 5280, 5692, 6503, 8385,
9730, 11167, 12389, 13946, 14003, 14086, 16339, 17887, 18058,
19329, 19436])}
Select Markers¶
After the filtering process, you can select the markers.
[10]:
baron_h1.sel_clust_marker()
-> Selecting markers for each cluster
You can view the selected markers as follows.
[11]:
baron_h1.get_markers()
[11]:
| Marker_1 | Marker_2 | Marker_3 | |
|---|---|---|---|
| C_acinar | PNLIPRP1 | CEL | CPA2 |
| C_activated_stellate | CRLF1 | SFRP2 | COL6A3 |
| C_alpha | IRX2 | GC | NPNT |
| C_beta | ADCYAP1 | HADH | G6PC2 |
| C_delta | LEPR | MIR7.3HG | PCP4 |
| C_ductal | MMP7 | KRT19 | TACSTD2 |
| C_endothelial | PLVAP | PECAM1 | CD93 |
| C_epsilon | GHRL | FRZB | NNMT |
| C_gamma | PPY | AQP3 | STMN2 |
| C_macrophage | PLA2G7 | C1QC | ITGB2 |
| C_mast | TPSB2 | TPSAB1 | CPA3 |
| C_quiescent_stellate | RGS5 | NDUFA4L2 | EDNRA |
| C_schwann | SOX10 | SEMA3C | GPM6B |
| C_t_cell | CD3D | ZAP70 | TRAC |
Or you can see the corresponding indexes in the data matrix.
[12]:
baron_h1.get_markers(ind_return = True)
[12]:
| Marker_1 | Marker_2 | Marker_3 | |
|---|---|---|---|
| C_acinar | 13082 | 3055 | 3733 |
| C_activated_stellate | 3836 | 15293 | 3637 |
| C_alpha | 8442 | 6534 | 11417 |
| C_beta | 322 | 7240 | 6383 |
| C_delta | 9345 | 10318 | 12504 |
| C_ductal | 10397 | 9048 | 16828 |
| C_endothelial | 13041 | 12627 | 2879 |
| C_epsilon | 6637 | 6299 | 11320 |
| C_gamma | 13386 | 932 | 16601 |
| C_macrophage | 12935 | 1973 | 8490 |
| C_mast | 17870 | 17869 | 3734 |
| C_quiescent_stellate | 14362 | 11057 | 4987 |
| C_schwann | 16160 | 15146 | 6911 |
| C_t_cell | 2842 | 19329 | 17887 |
You can also pull selected markers from the data matrix.
[13]:
mark_data = baron_h1.get_marker_data()
mark_data
[13]:
{'acinar': array([[4.9068906 , 3.5849625 , 6.89481776],
[2.80735492, 3.169925 , 6.82017896],
[4.39231742, 4.70043972, 5.4918531 ],
...,
[0. , 0. , 0. ],
[0. , 0. , 0. ],
[0. , 0. , 0. ]]),
'activated_stellate': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
'alpha': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
'beta': array([[0., 2., 0.],
[0., 1., 0.],
[0., 2., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
'delta': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 1., 1.],
[0., 0., 0.]]),
'ductal': array([[0. , 0. , 0. ],
[0. , 2.80735492, 4.24792751],
[0. , 0. , 2.5849625 ],
...,
[0. , 0. , 0. ],
[0. , 0. , 0. ],
[0. , 0. , 0. ]]),
'endothelial': array([[0. , 0. , 0. ],
[0. , 0. , 0. ],
[0. , 0. , 0. ],
...,
[3.169925, 2. , 3. ],
[0. , 0. , 0. ],
[0. , 0. , 0. ]]),
'epsilon': array([[0., 0., 0.],
[0., 0., 0.],
[1., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
'gamma': array([[2.32192809, 0. , 0. ],
[1. , 1.5849625 , 0. ],
[1.5849625 , 0. , 0. ],
...,
[0. , 0. , 0. ],
[0. , 0. , 1. ],
[2. , 0. , 0. ]]),
'macrophage': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
'mast': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
'quiescent_stellate': array([[0. , 0. , 0. ],
[0. , 0. , 0. ],
[0. , 0. , 0. ],
...,
[0. , 0. , 0. ],
[0. , 0. , 0. ],
[2.32192809, 1.5849625 , 1. ]]),
'schwann': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]),
't_cell': array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]])}
5 markers are selected by default for each cluster. You can access the selected markers for each cluster with the dictionary keys.
[14]:
schwann = mark_data['schwann']
schwann
[14]:
array([[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.],
...,
[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]])
[15]:
schwann.shape
[15]:
(1937, 3)
You can also perform marker selection with dynamic programming.
[16]:
baron_h1.sel_clust_marker(dyn_prog=True)
-> Selecting markers for each cluster
-> |⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛| 100% Number of Clusters With Selected Markers : 14
If you want, you can increase the number of markers to be selected.
Note
If you are going to increase the number of markers, make sure that the number of genes remaining after filtering is more than the number of markers to be selected.
[17]:
baron_h1.filter_genes(nof_sel=20)
baron_h1.sel_clust_marker(nof_markers=10)
-> Eliminating low expression genes
-> Selecting cluster-specific candidate marker genes
-> Selecting markers for each cluster
[18]:
baron_h1.get_markers(ind_return = True)
[18]:
| Marker_1 | Marker_2 | Marker_3 | Marker_4 | Marker_5 | Marker_6 | Marker_7 | Marker_8 | Marker_9 | Marker_10 | |
|---|---|---|---|---|---|---|---|---|---|---|
| C_acinar | 13082 | 3055 | 3733 | 8968 | 16742 | 4020 | 3057 | 12577 | 13083 | 13081 |
| C_activated_stellate | 3836 | 15293 | 3637 | 18923 | 17233 | 6179 | 5878 | 3632 | 9231 | 405 |
| C_alpha | 8442 | 6534 | 11417 | 10734 | 2390 | 8713 | 14361 | 17366 | 15991 | 12291 |
| C_beta | 322 | 7240 | 6383 | 5342 | 11265 | 68 | 13322 | 1974 | 15686 | 4675 |
| C_delta | 9345 | 10318 | 12504 | 1449 | 14210 | 6407 | 4516 | 6543 | 959 | 15098 |
| C_ductal | 10397 | 9048 | 16828 | 3383 | 13242 | 9077 | 13639 | 17311 | 4203 | 809 |
| C_endothelial | 13041 | 12627 | 2879 | 12396 | 6159 | 8756 | 16167 | 13115 | 17277 | 4968 |
| C_epsilon | 6637 | 6299 | 11320 | 16388 | 19024 | 9723 | 6358 | 12939 | 6027 | 190 |
| C_gamma | 13386 | 932 | 16601 | 7869 | 10138 | 18672 | 12375 | 18319 | 11391 | 2159 |
| C_macrophage | 12935 | 1973 | 8490 | 15095 | 1971 | 3888 | 7541 | 4127 | 15487 | 7307 |
| C_mast | 17870 | 17869 | 3734 | 2816 | 14929 | 14396 | 2020 | 6502 | 8896 | 9696 |
| C_quiescent_stellate | 14362 | 11057 | 4987 | 8266 | 6668 | 315 | 19198 | 372 | 3930 | 1823 |
| C_schwann | 16160 | 15146 | 6911 | 12598 | 3859 | 1728 | 16509 | 6356 | 4579 | 7961 |
| C_t_cell | 2842 | 19329 | 17887 | 2729 | 18058 | 6503 | 8385 | 12389 | 2840 | 2854 |