Hugging Face's logo

scvi-tools
/
tabula-sapiens-bladder-scanvi

TensorBoard
scvi-tools
biology
genomics
single-cell
model_cls_name:SCANVI
scvi_version:1.4.2
anndata_version:0.12.7
modality:rna
tissue:various
annotated:True
Model card Files
xet
 Metrics Community
OriKron commited on
Commit
6579398
·
verified ·
1 Parent(s): 1322bbf

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +56 -47
README.md CHANGED
@@ -6,8 +6,8 @@ tags:
6
  - genomics
7
  - single-cell
8
  - model_cls_name:SCANVI
9
- - scvi_version:1.2.0
10
- - anndata_version:0.11.1
11
  - modality:rna
12
  - tissue:various
13
  - annotated:True
@@ -23,7 +23,7 @@ clustering.
23
 
24
  scANVI takes as input a scRNA-seq gene expression matrix with cells and genes as well as a
25
  cell-type annotation for a subset of cells.
26
- We provide an extensive [user guide](https://docs.scvi-tools.org/en/1.2.0/user_guide/models/scanvi.html).
27
 
28
  - See our original manuscript for further details of the model:
29
  [scANVI manuscript](https://www.embopress.org/doi/full/10.15252/msb.20209620).
@@ -31,7 +31,7 @@ We provide an extensive [user guide](https://docs.scvi-tools.org/en/1.2.0/user_g
31
  how to leverage pre-trained models.
32
 
33
  This model can be used for fine tuning on new data using our Arches framework:
34
- [Arches tutorial](https://docs.scvi-tools.org/en/1.0.0/tutorials/notebooks/scarches_scvi_tools.html).
35
 
36
 
37
  # Model Description
@@ -54,10 +54,10 @@ space might still be useful for analysis.
54
 
55
  | Metric | Training Value | Validation Value |
56
  |-------------------------|----------------|------------------|
57
- | Mean Absolute Error | 1.77 | 1.80 |
58
- | Pearson Correlation | 0.87 | 0.88 |
59
- | Spearman Correlation | 0.79 | 0.77 |
60
- | R² (R-Squared) | 0.52 | 0.51 |
61
 
62
  The gene-wise coefficient of variation summarizes how well variation between different genes is
63
  preserved by the generated model expression. This value is usually quite high.
@@ -66,10 +66,10 @@ preserved by the generated model expression. This value is usually quite high.
66
 
67
  | Metric | Training Value |
68
  |-------------------------|----------------|
69
- | Mean Absolute Error | 16.06 |
70
- | Pearson Correlation | 0.73 |
71
- | Spearman Correlation | 0.77 |
72
- | R² (R-Squared) | 0.08 |
73
 
74
  </details>
75
 
@@ -86,21 +86,28 @@ cell-type.
86
 
87
  | Index | gene_f1 | lfc_mae | lfc_pearson | lfc_spearman | roc_auc | pr_auc | n_cells |
88
  | --- | --- | --- | --- | --- | --- | --- | --- |
89
- | fibroblast | 0.97 | 1.03 | 0.77 | 0.96 | 0.34 | 0.90 | 5557.00 |
90
- | macrophage | 0.94 | 1.05 | 0.73 | 0.95 | 0.30 | 0.88 | 5338.00 |
91
- | bladder urothelial cell | 0.93 | 0.89 | 0.80 | 0.97 | 0.42 | 0.91 | 4151.00 |
92
- | T cell | 0.95 | 1.86 | 0.74 | 0.90 | 0.24 | 0.86 | 2916.00 |
93
- | myofibroblast cell | 0.93 | 1.88 | 0.66 | 0.88 | 0.34 | 0.84 | 2078.00 |
94
- | plasma cell | 0.88 | 1.87 | 0.72 | 0.89 | 0.14 | 0.86 | 1141.00 |
95
- | mast cell | 0.93 | 2.78 | 0.61 | 0.82 | 0.21 | 0.79 | 1029.00 |
96
- | pericyte | 0.93 | 2.05 | 0.74 | 0.88 | 0.27 | 0.78 | 875.00 |
97
- | mature NK T cell | 0.86 | 3.00 | 0.67 | 0.82 | 0.42 | 0.87 | 508.00 |
98
- | smooth muscle cell | 0.94 | 2.88 | 0.72 | 0.81 | 0.28 | 0.82 | 290.00 |
99
- | vein endothelial cell | 0.79 | 3.02 | 0.70 | 0.86 | 0.36 | 0.79 | 278.00 |
100
- | B cell | 0.87 | 3.88 | 0.58 | 0.70 | 0.34 | 0.75 | 253.00 |
101
- | capillary endothelial cell | 0.72 | 3.22 | 0.71 | 0.76 | 0.38 | 0.75 | 77.00 |
102
- | endothelial cell of lymphatic vessel | 0.75 | 4.50 | 0.66 | 0.73 | 0.28 | 0.70 | 74.00 |
103
- | plasmacytoid dendritic cell | 0.62 | 5.99 | 0.53 | 0.47 | 0.32 | 0.73 | 18.00 |
 
 
 
 
 
 
 
104
 
105
  </details>
106
 
@@ -120,7 +127,9 @@ These provide the settings to setup the original model:
120
  "dropout_rate": 0.05,
121
  "dispersion": "gene",
122
  "gene_likelihood": "nb",
 
123
  "linear_classifier": false,
 
124
  "latent_distribution": "normal",
125
  "use_batch_norm": "none",
126
  "use_layer_norm": "both",
@@ -136,9 +145,9 @@ These provide the settings to setup the original model:
136
  Arguments passed to setup_anndata of the original model:
137
  ```json
138
  {
139
- "labels_key": "cell_ontology_class",
140
  "unlabeled_category": "unknown",
141
- "layer": null,
142
  "batch_key": "donor_assay",
143
  "size_factor_key": null,
144
  "categorical_covariate_keys": null,
@@ -153,15 +162,15 @@ Arguments passed to setup_anndata of the original model:
153
  <summary><strong>Data Registry</strong></summary>
154
 
155
  Registry elements for AnnData manager:
156
- | Registry Key | scvi-tools Location |
157
- |-------------------|--------------------------------------|
158
- | X | adata.X |
159
- | batch | adata.obs['_scvi_batch'] |
160
- | labels | adata.obs['_scvi_labels'] |
161
- | latent_qzm | adata.obsm['scanvi_latent_qzm'] |
162
- | latent_qzv | adata.obsm['scanvi_latent_qzv'] |
163
- | minify_type | adata.uns['_scvi_adata_minify_type'] |
164
- | observed_lib_size | adata.obs['observed_lib_size'] |
165
 
166
  - **Data is Minified**: False
167
 
@@ -170,16 +179,16 @@ Registry elements for AnnData manager:
170
  <details>
171
  <summary><strong>Summary Statistics</strong></summary>
172
 
173
- | Summary Stat Key | Value |
174
  |--------------------------|-------|
175
- | n_batch | 5 |
176
- | n_cells | 24583 |
177
- | n_extra_categorical_covs | 0 |
178
- | n_extra_continuous_covs | 0 |
179
- | n_labels | 16 |
180
- | n_latent_qzm | 20 |
181
- | n_latent_qzv | 20 |
182
- | n_vars | 3000 |
183
 
184
  </details>
185
 
 
6
  - genomics
7
  - single-cell
8
  - model_cls_name:SCANVI
9
+ - scvi_version:1.4.2
10
+ - anndata_version:0.12.7
11
  - modality:rna
12
  - tissue:various
13
  - annotated:True
 
23
 
24
  scANVI takes as input a scRNA-seq gene expression matrix with cells and genes as well as a
25
  cell-type annotation for a subset of cells.
26
+ We provide an extensive [user guide](https://docs.scvi-tools.org/en/stable/user_guide/models/scanvi.html).
27
 
28
  - See our original manuscript for further details of the model:
29
  [scANVI manuscript](https://www.embopress.org/doi/full/10.15252/msb.20209620).
 
31
  how to leverage pre-trained models.
32
 
33
  This model can be used for fine tuning on new data using our Arches framework:
34
+ [Arches tutorial](https://docs.scvi-tools.org/en/stable/tutorials/notebooks/scrna/scarches_scvi_tools.html).
35
 
36
 
37
  # Model Description
 
54
 
55
  | Metric | Training Value | Validation Value |
56
  |-------------------------|----------------|------------------|
57
+ | Mean Absolute Error | 1.37 | 1.40 |
58
+ | Pearson Correlation | 0.86 | 0.86 |
59
+ | Spearman Correlation | 0.79 | 0.79 |
60
+ | R² (R-Squared) | 0.59 | 0.60 |
61
 
62
  The gene-wise coefficient of variation summarizes how well variation between different genes is
63
  preserved by the generated model expression. This value is usually quite high.
 
66
 
67
  | Metric | Training Value |
68
  |-------------------------|----------------|
69
+ | Mean Absolute Error | 22.95 |
70
+ | Pearson Correlation | 0.74 |
71
+ | Spearman Correlation | 0.76 |
72
+ | R² (R-Squared) | 0.27 |
73
 
74
  </details>
75
 
 
86
 
87
  | Index | gene_f1 | lfc_mae | lfc_pearson | lfc_spearman | roc_auc | pr_auc | n_cells |
88
  | --- | --- | --- | --- | --- | --- | --- | --- |
89
+ | bladder urothelial cell | 0.94 | 0.34 | 0.92 | 0.99 | 0.07 | 0.02 | 27241.00 |
90
+ | fibroblast | 0.95 | 0.39 | 0.86 | 0.98 | 0.06 | 0.02 | 13766.00 |
91
+ | monocyte | 0.87 | 0.56 | 0.76 | 0.96 | 0.05 | 0.02 | 4855.00 |
92
+ | CD8-positive, alpha-beta T cell | 0.89 | 1.16 | 0.62 | 0.91 | 0.08 | 0.02 | 4432.00 |
93
+ | macrophage | 0.92 | 0.89 | 0.69 | 0.94 | 0.04 | 0.02 | 3152.00 |
94
+ | myofibroblast cell | 0.88 | 0.74 | 0.76 | 0.94 | 0.37 | 0.17 | 3136.00 |
95
+ | CD4-positive, alpha-beta T cell | 0.90 | 1.35 | 0.65 | 0.91 | 0.04 | 0.02 | 2189.00 |
96
+ | mast cell | 0.96 | 1.83 | 0.65 | 0.84 | 0.04 | 0.02 | 1568.00 |
97
+ | plasma cell | 0.92 | 1.25 | 0.74 | 0.91 | 0.08 | 0.02 | 1205.00 |
98
+ | pericyte | 0.88 | 1.40 | 0.71 | 0.88 | 0.14 | 0.07 | 1184.00 |
99
+ | smooth muscle cell | 0.92 | 1.51 | 0.72 | 0.89 | 0.10 | 0.03 | 1053.00 |
100
+ | T cell | 0.87 | 2.11 | 0.61 | 0.83 | 0.14 | 0.08 | 781.00 |
101
+ | B cell | 0.87 | 2.77 | 0.52 | 0.78 | 0.12 | 0.04 | 484.00 |
102
+ | vein endothelial cell | 0.76 | 1.98 | 0.67 | 0.85 | 0.11 | 0.03 | 438.00 |
103
+ | neutrophil | 0.89 | 4.58 | 0.50 | 0.60 | 0.24 | 0.02 | 241.00 |
104
+ | natural killer cell | 0.79 | 4.36 | 0.53 | 0.69 | 0.22 | 0.02 | 165.00 |
105
+ | endothelial cell of lymphatic vessel | 0.74 | 4.12 | 0.55 | 0.66 | 0.20 | 0.02 | 127.00 |
106
+ | capillary endothelial cell | 0.74 | 3.20 | 0.61 | 0.71 | 0.18 | 0.02 | 107.00 |
107
+ | erythrocyte | 0.24 | 6.19 | 0.39 | 0.27 | 0.39 | 0.03 | 106.00 |
108
+ | mature NK T cell | 0.69 | 5.02 | 0.57 | 0.60 | 0.32 | 0.02 | 95.00 |
109
+ | endothelial cell | 0.51 | 4.22 | 0.61 | 0.71 | 0.20 | 0.02 | 37.00 |
110
+ | regulatory T cell | 0.61 | 5.24 | 0.61 | 0.64 | 0.30 | 0.02 | 23.00 |
111
 
112
  </details>
113
 
 
127
  "dropout_rate": 0.05,
128
  "dispersion": "gene",
129
  "gene_likelihood": "nb",
130
+ "use_observed_lib_size": true,
131
  "linear_classifier": false,
132
+ "datamodule": null,
133
  "latent_distribution": "normal",
134
  "use_batch_norm": "none",
135
  "use_layer_norm": "both",
 
145
  Arguments passed to setup_anndata of the original model:
146
  ```json
147
  {
148
+ "labels_key": "cell_type",
149
  "unlabeled_category": "unknown",
150
+ "layer": "counts",
151
  "batch_key": "donor_assay",
152
  "size_factor_key": null,
153
  "categorical_covariate_keys": null,
 
162
  <summary><strong>Data Registry</strong></summary>
163
 
164
  Registry elements for AnnData manager:
165
+ | Registry Key | scvi-tools Location |
166
+ |--------------------------|--------------------------------------|
167
+ | X | adata.layers['counts'] |
168
+ | batch | adata.obs['_scvi_batch'] |
169
+ | labels | adata.obs['_scvi_labels'] |
170
+ | latent_qzm | adata.obsm['scanvi_latent_qzm'] |
171
+ | latent_qzv | adata.obsm['scanvi_latent_qzv'] |
172
+ | minify_type | adata.uns['_scvi_adata_minify_type'] |
173
+ | observed_lib_size | adata.obs['observed_lib_size'] |
174
 
175
  - **Data is Minified**: False
176
 
 
179
  <details>
180
  <summary><strong>Summary Statistics</strong></summary>
181
 
182
+ | Summary Stat Key | Value |
183
  |--------------------------|-------|
184
+ | n_batch | 9 |
185
+ | n_cells | 66385 |
186
+ | n_extra_categorical_covs | 0 |
187
+ | n_extra_continuous_covs | 0 |
188
+ | n_labels | 23 |
189
+ | n_latent_qzm | 20 |
190
+ | n_latent_qzv | 20 |
191
+ | n_vars | 3000 |
192
 
193
  </details>
194