exprmat 0.1.46__tar.gz → 0.1.47__tar.gz

{exprmat-0.1.46 → exprmat-0.1.47}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: exprmat
-Version: 0.1.46
+Version: 0.1.47
 Classifier: Programming Language :: Rust
 Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/clustering/nmf.py

@@ -2,6 +2,7 @@
 from scipy.cluster.hierarchy import cophenet, single
 from scipy.spatial.distance import pdist, squareform
 from sklearn.decomposition import non_negative_factorization
+import numpy as np
 
 
 def nmf(
@@ -12,6 +13,7 @@ def nmf(
     random_state = 42, verbose = 0, shuffle = False
 ):
     
+    matrix[np.isnan(matrix)] = 0
     w, h, n = non_negative_factorization(
         matrix, W, H, n_components = n_comps,
         init = init, 

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/clustering/seriation.py

@@ -33,7 +33,7 @@ def seriation(Z, N, cur_index):
 
 def serial_matrix(dist, method = "ward", seed = 42):
     '''
-    Compute_serial_matrix transforms a distance matrix into a sorted distance matrix 
+    Transforms a distance matrix into a sorted distance matrix 
     according to the order implied by the hierarchical tree (dendrogram).
 
     Parameters
@@ -55,7 +55,7 @@ def serial_matrix(dist, method = "ward", seed = 42):
 
     N = len(dist)
     flat_dist_mat = squareform(dist)
-    res_linkage = linkage(flat_dist_mat, method = method,preserve_input = True)
+    res_linkage = linkage(flat_dist_mat, method = method, preserve_input = True)
     res_order = seriation(res_linkage, N, N + N-2)
     seriated_dist = np.zeros((N,N))
     a,b = np.triu_indices(N, k = 1)

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/plotting/cnmf.py

@@ -73,6 +73,7 @@ def cnmf_distance_usages(
         elif annotations in adata.obsm.keys(): 
             annotations = adata[subset, :][order, :].obsm[annotations]
 
+    print(indices)
     return matrix_plot(
         serial, labels = indices, annotations = annotations, cmap_annotations = cmap_annotations, 
         cmap = cmap, figsize = figsize, dpi = dpi, show_indices = show_indices,
@@ -103,6 +104,7 @@ def cnmf_distance_modules(
         elif annotations in adata.varm.keys(): 
             annotations = adata[:, subset][:, order].varm[annotations]
 
+    print(indices)
     return matrix_plot(
         serial, labels = indices, annotations = annotations, cmap_annotations = cmap_annotations, 
         cmap = cmap, figsize = figsize, dpi = dpi, show_indices = show_indices, legend_cols = legend_cols

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/reader/experiment.py

@@ -41,7 +41,7 @@ class experiment:
         dump = '.', subset = None,
         version = em.SPECIFICATION
     ):
-        # TODO: we support rna only at present.
+        
         table = meta.dataframe.to_dict(orient = 'list')
         self.mudata = mudata
         self.modalities = modalities
@@ -1694,6 +1694,19 @@ class experiment:
             self.modalities['rna'][props['sample']] = mc
 
     
+    def run_rna_construct_atlas(self, run_on_samples = False, **kwargs):
+        self.do_for_rna(
+            run_on_samples, st.rna_construct_atlas, 
+            expm_dir = self.directory,
+            expm_subset = self.subset,
+            **kwargs
+        )
+
+    
+    def run_rna_project(self, run_on_samples = False, **kwargs):
+        self.do_for_rna(run_on_samples, st.rna_project, **kwargs)
+
+    
     def run_atac_make_bins(self, run_on_samples = False, **kwargs):
         # bins can be made directly from bedgraph files if there is no alignment bam provided.
         # thus these samples can be used for clustering.
@@ -1998,6 +2011,9 @@ class experiment:
     def plot_rna_kde(self, run_on_samples = False, **kwargs):
         return self.plot_for_rna(run_on_samples, st.rna_plot_kde, **kwargs)
     
+    def plot_rna_projection(self, run_on_samples = False, **kwargs):
+        return self.plot_for_rna(run_on_samples, st.rna_plot_projection, **kwargs)
+    
     def plot_rna_proportion(self, run_on_samples = False, **kwargs):
         return self.plot_for_rna(run_on_samples, st.rna_plot_proportion, **kwargs)
     

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/reader/static.py

@@ -1015,6 +1015,22 @@ def rna_metacell(
     return metacell[0]
 
 
+def rna_construct_atlas(
+    adata, sample_names, expm_dir, expm_subset, **kwargs
+):
+    from exprmat.transfer.atlas import construct_atlas
+    if sample_names != 'integrated':
+        expm_dir = os.path.join(expm_dir, 'rna')
+        expm_subset = sample_names
+    construct_atlas(adata, expm_dir, expm_subset, **kwargs)
+
+
+def rna_project(
+    adata, sample_names, **kwargs
+):
+    from exprmat.transfer.atlas import project
+    project(expm = adata, **kwargs)
+
 def atac_filter_cells(
     adata, sample_name, min_counts = 5000, max_counts = 100000, 
     min_tsse = 10, max_tsse = 100
@@ -1713,6 +1729,54 @@ def rna_plot_kde(
     return fig
 
 
+def rna_plot_projection(
+    # query projection
+    adata, sample_name, color,
+    figsize = (3, 3), dpi = 100, projection_key = 'projection', 
+
+    # atlas background
+    background = 'annotation',
+    annotate = True, annotate_fontsize = 9, 
+    **kwargs
+):
+    from exprmat.reduction.plot import embedding
+    import matplotlib.pyplot as plt
+    import warnings
+    warnings.filterwarnings('ignore')
+
+    fig, axes = plt.subplots(1, 1, dpi = dpi, figsize = figsize)
+
+    from scipy.sparse import csr_matrix
+    tab = adata.uns[projection_key]
+    nbg = len(tab)
+    bg = ad.AnnData(csr_matrix((nbg, 2)))
+    bg.obs['annotation'] = tab['annotation'].tolist()
+    bg.obs['batch'] = tab['batch'].tolist()
+    bg.obs['annotation'] = bg.obs['annotation'].astype('category')
+    bg.obs['batch'] = bg.obs['batch'].astype('category')
+    bg.obsm['basis'] = tab[['xf', 'yf']].copy().values
+
+    embedding(
+        bg, basis = 'basis', color = background,
+        default_color = '#eeeeee', ptsize = 8,
+        annotate = annotate, annotate_style = 'text',
+        legend = False, contour_plot = False,
+        annotate_fontsize = annotate_fontsize,
+        annotate_only = None,
+        cmap = None, add_outline = True, outline_color = '#777777',
+        ax = axes, dpi = dpi,
+        sample_name = sample_name
+    )
+    
+    embedding(
+        adata, basis = projection_key, color = color,
+        ax = axes, dpi = dpi, 
+        sample_name = sample_name, **kwargs
+    )
+
+    return fig
+
+
 def rna_plot_dot(
     adata, sample_name, figsize, dpi, 
     var_names, groupby, *, log = False,

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/reduction/mde.py

@@ -63,7 +63,10 @@ def mde(
 
 
 def mde_fit(
-    data, fit, fit_mask, device = None, **kwargs
+    data, fit, fit_mask, device = None, 
+    eps: float = 0.00001, max_iter: int = 300, memory_size: int = 10,
+    verbose = False,
+    **kwargs
 ):
     
     try: import pymde
@@ -90,8 +93,12 @@ def mde_fit(
             torch.nonzero(torch.tensor(fit_mask)).squeeze(), 
             torch.tensor(fitting, dtype = torch.float)
         ),
-        verbose = True
-    ).embed(verbose = True)
+        verbose = verbose
+    ).embed(
+        verbose = verbose,
+        eps = eps, max_iter = max_iter, 
+        memory_size = memory_size
+    )
 
     if isinstance(emb, torch.Tensor):
         emb = emb.cpu().numpy()

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/reduction/plot.py

@@ -301,7 +301,7 @@ def embedding(
     legend = True,
 
     # contour plotting option.
-    contour_plot = True,
+    contour_plot = False,
     contour_fill = False,
     contour_mask = None,
     contour_mask_values = [],
@@ -426,22 +426,6 @@ def embedding(
             hue_order = hue_order if default_palette is not None else None
         )
 
-        if contour_plot:
-
-            cx = atlas_data['x']
-            cy = atlas_data['y']
-            if contour_mask is not None:
-                cx = [y for x, y in zip(adata.obs[contour_mask], cx) if x in contour_mask_values]
-                cy = [y for x, y in zip(adata.obs[contour_mask], cy) if x in contour_mask_values]
-
-            sb.kdeplot(
-                x = cx, y = cy, warn_singular = False,
-                linewidths = contour_linewidth, bw_adjust = contour_bw, bw_method = 'scott',
-                fill = contour_fill, ax = axes, 
-                palette = None, color = contour_default_color, alpha = contour_alpha,
-                levels = contour_levels, legend = False
-            )
-
         if legend:
             assert len(adata.uns[f'{color}.colors']) == len(hue_order)
             dummy_objects = []
@@ -492,6 +476,7 @@ def embedding(
                 axes.add_artist(text)
                 pass
     
+    
     elif type(labels[0]) is bool:
 
         sb.scatterplot(
@@ -546,7 +531,24 @@ def embedding(
     
     else:
         warning('color label must be categorical (string) or numerical.')
-        return None
+        return 
+
+
+    if contour_plot:
+
+        cx = atlas_data['x']
+        cy = atlas_data['y']
+        if contour_mask is not None:
+            cx = [y for x, y in zip(adata.obs[contour_mask], cx) if x in contour_mask_values]
+            cy = [y for x, y in zip(adata.obs[contour_mask], cy) if x in contour_mask_values]
+
+        sb.kdeplot(
+            x = cx, y = cy, warn_singular = False,
+            linewidths = contour_linewidth, bw_adjust = contour_bw, bw_method = 'scott',
+            fill = contour_fill, ax = axes, 
+            palette = None, color = contour_default_color, alpha = contour_alpha,
+            levels = contour_levels, legend = False
+        )
     
     def format_coord(x, y): return f"x = {x:.2f}, y = {y:.2f}"
     axes.format_coord = format_coord

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/trajectory/trace.py

@@ -86,6 +86,7 @@ def binned_statistics(sorted_pstime, trajectory_key = 'ppt', statistic = 'mean',
 
     X = sorted_pstime.X.T
     if issparse(X): X = X.todense()
+    X = np.array(X)
     minx = sorted_pstime.obs[f'{trajectory_key}.pseudotime'].min()
     maxx = sorted_pstime.obs[f'{trajectory_key}.pseudotime'].max()
 

{exprmat-0.1.46 → exprmat-0.1.47}/exprmat/transfer/atlas.py

@@ -8,7 +8,9 @@ from exprmat.reader.experiment import experiment
 
 
 def construct_atlas(
-    expm: experiment, 
+    adata,
+    expm_dir,
+    expm_subset,
     key_counts = 'counts',
     key_batch = 'batch',
     hvg = 'hvg',
@@ -119,14 +121,14 @@ def construct_atlas(
     warnings.filterwarnings('ignore')
     torch.set_float32_matmul_precision('high')
 
-    atlas = expm.mudata['rna']
-    savename = expm.subset if expm.subset is not None else 'integrated'
-    path = os.path.join(expm.directory, 'scvi', savename)
-    scanvipath = os.path.join(expm.directory, 'scanvi', savename)
+    atlas = adata
+    savename = expm_subset if expm_subset is not None else 'integrated'
+    path = os.path.join(expm_dir, 'scvi', savename)
+    scanvipath = os.path.join(expm_dir, 'scanvi', savename)
     
     if os.path.exists(path):
         error('the destination path `scvi/{savename}` is already occupied.')
-    else: os.makedirs(os.path.join(expm.directory, 'scvi'), exist_ok = True)
+    else: os.makedirs(os.path.join(expm_dir, 'scvi'), exist_ok = True)
 
 
     # extract count matrix.
@@ -234,32 +236,30 @@ def construct_atlas(
     info(f'trained scanvi model saved.')
     scanvi_model.save(scanvipath)
     info(f'saving metadata of atlas as a light version.')
-    metadata.write_h5ad(os.path.join(expm.directory, 'scvi', savename, 'metadata.h5ad'))
+    metadata.write_h5ad(os.path.join(expm_dir, 'scvi', savename, 'metadata.h5ad'))
     warnings.filterwarnings('default')
 
 
-def is_atlas(expm: experiment):
+def is_atlas(adata):
 
-    expm.check_merged('rna')
-
-    if 'atlas' not in expm.mudata['rna'].uns.keys():
+    if 'atlas' not in adata.uns.keys():
         warning(f'do not contain an `atlas` key in uns slot. build atlas first.')
         return False
     
-    atlas = expm.mudata['rna'].uns['atlas']
-    if atlas['batch'] not in expm.mudata['rna'].obs.keys():
+    atlas = adata.uns['atlas']
+    if atlas['batch'] not in adata.obs.keys():
         warning(f'`{atlas["batch"]}` do not exist.')
         return False
     
-    if atlas['annotation'] not in expm.mudata['rna'].obs.keys():
+    if atlas['annotation'] not in adata.obs.keys():
         warning(f'`{atlas["annotation"]}` do not exist.')
         return False
 
-    if atlas['scvi'] not in expm.mudata['rna'].obsm.keys():
+    if atlas['scvi'] not in adata.obsm.keys():
         warning(f'`{atlas["scvi"]}` do not exist.')
         return False
 
-    if atlas['scanvi'] not in expm.mudata['rna'].obsm.keys():
+    if atlas['scanvi'] not in adata.obsm.keys():
         warning(f'`{atlas["scanvi"]}` do not exist.')
         return False
 
@@ -268,7 +268,7 @@ def is_atlas(expm: experiment):
 
 def embed(
     atlas: experiment, 
-    expm: experiment,
+    expm,
     
     counts_key = 'counts',
     batch_key = 'batch',
@@ -281,7 +281,7 @@ def embed(
     import scvi
     import anndata
 
-    input = expm.mudata['rna']
+    input = expm
     query = anndata.AnnData(
         X = input.X.copy() if counts_key is None 
         else input.layers[counts_key].copy()
@@ -326,4 +326,75 @@ def embed(
         query.obs[predict_labels] = query_model.predict()
         input.obs[predict_labels] = query.obs[predict_labels]
     
-    return query
+    return query
+
+
+def project(
+    atlas: experiment, 
+    expm,
+    
+    atlas_modality = 'rna',
+    atlas_embedding = 'umap',
+    counts_key = 'counts',
+    batch_key = 'batch',
+    key_query_latent = 'scanvi',
+    scanvi_unlabel = 'unknown',
+    scvi_epoch_reduction = 3,
+    predict_labels = 'annot',
+    key_added = 'projection',
+
+    re_embed = False,
+    **kwargs
+):
+    
+    data_meta = atlas[atlas_modality].obs[[
+        atlas[atlas_modality].uns['atlas']['batch'],
+        atlas[atlas_modality].uns['atlas']['annotation']
+    ]].copy()
+    data_meta.columns = ['batch', 'annotation']
+
+    embed(
+        atlas = atlas,
+        expm = expm,
+        counts_key = counts_key,
+        batch_key = batch_key,
+        key_query_latent = key_query_latent,
+        scanvi_unlabel = scanvi_unlabel,
+        scvi_epoch_reduction = scvi_epoch_reduction,
+        predict_labels = predict_labels
+    )
+
+    # copy metadata from atlas to expm.
+    
+    atlas.check_merged(atlas_modality)
+    atlas = atlas[atlas_modality]
+    expm = expm
+
+    data = atlas.obsm[atlas_embedding]
+
+    data_meta['x'] = data[:, 0]
+    data_meta['y'] = data[:, 1]
+
+    import pandas as pd
+    if isinstance(data, pd.DataFrame): data = data.values
+    if not isinstance(data, np.ndarray): data = np.array(data)
+    assert data.shape[1] == 2
+
+    n_cells = expm.n_obs
+    n_ref = atlas.n_obs
+
+    scanvi_atlas = atlas.obsm[atlas.uns['atlas']['scanvi']]
+    scanvi_expm = expm.obsm[key_query_latent]
+
+    merged_latent = np.concat([scanvi_atlas, scanvi_expm], axis = 0)
+    merged_embed = np.concat([data, np.zeros(shape = (n_cells, 2))], axis = 0)
+    mask = np.array([True] * n_ref + [False] * n_cells)
+
+    from exprmat.reduction.mde import mde_fit, mde
+    if not re_embed: emb = mde_fit(merged_latent, merged_embed, mask, **kwargs)
+    else: emb = mde(merged_latent, **kwargs)
+
+    expm.obsm[key_added] = emb[~mask, :]
+    data_meta['xf'] = emb[mask, 0]
+    data_meta['yf'] = emb[mask, 1]
+    expm.uns[key_added] = data_meta

{exprmat-0.1.46 → exprmat-0.1.47}/pyproject.toml

@@ -20,7 +20,7 @@ description = "exprmat: Routines for expression matrices"
 readme = "readme.md"
 license = {file = "license"}
 requires-python = ">=3.9, <3.13"
-version = "0.1.46"
+version = "0.1.47"
 authors = [
     {name = "Zheng Yang", email = "xornent@outlook.com"}
 ]