Merge branch 'main' into fc_int2

.github/workflows/integration01-ci.yml

@@ -73,6 +73,7 @@ jobs:
         run: |
           cd teaseq/integration
           curl -o pipeline.yml https://raw.githubusercontent.com/DendrouLab/panpipes/main/tests/integration_1/pipeline.yml  
+
       - name: Replace template contents in configuration file
         run: |
           cd teaseq/integration

panpipes/python_scripts/batch_correct_multivi.py

@@ -57,7 +57,7 @@
 sc.settings.autoshow = False
 sc.settings.figdir = args.figdir
 
-
+scvi.settings.seed = 1492
 # load parameters
 
 threads_available = multiprocessing.cpu_count()
@@ -71,17 +71,19 @@
 # ------------------------------------------------------------------
 L.info("Running multivi")
 
-# mdata = mu.read(args.scaled_anndata)
-# rna = mdata['rna'].copy()
-# atac = mdata['atac'].copy()
-rna= mu.read(args.scaled_anndata +"/" + "rna")
-atac= mu.read(args.scaled_anndata +"/" + "atac")
+mdata = mu.read(args.scaled_anndata)
+rna = mdata['rna'].copy()
+atac = mdata['atac'].copy()
 
+del mdata
 
 if check_for_bool(params["multimodal"]["MultiVI"]["lowmem"]):
-    L.info("subsetting atac to top 25k HVF")
+    if 'hvg' in atac.uns.keys():
+        L.info("subsetting atac to top HVF")
+        atac = atac[:, atac.var.highly_variable].copy()
+    L.info("calculating and subsetting atac to top 25k HVF")
     sc.pp.highly_variable_genes(atac, n_top_genes=25000)
-    atac = atac[:, atac.var.highly_variable]
+    atac = atac[:, atac.var.highly_variable].copy()
 
 
 
@@ -126,6 +128,15 @@
 L.info("concatenating modalities to comply with multiVI")
 # adata_paired = ad.concat([rna, atac], join="outer")
 # adata_paired.var = pd.concat([rna.var,atac.var])
+if rna.is_view:
+    L.info("RNA is view")
+    atac = rna.copy()
+if atac.is_view:
+    L.info("ATAC is view")
+    atac = atac.copy()
+
+L.info(atac.is_view)
+
 
 adata_paired = ad.concat([rna.T, atac.T]).T
 
@@ -223,14 +234,14 @@
     multivi_training_args={}
 else:
     multivi_training_args =  {k: v for k, v in params["multimodal"]['MultiVI']['training_args'].items() if v is not None}
-
+L.info("multivi training args")
 print(multivi_training_args)
 
 if params["multimodal"]['MultiVI']['training_plan'] is None:
     multivi_training_plan = {}
 else:
     multivi_training_plan =  {k: v for k, v in params["multimodal"]['MultiVI']['training_plan'].items() if v is not None}
-
+L.info("multivi training plan")
 print(multivi_training_plan)
 
 mvi.view_anndata_setup()

panpipes/python_scripts/plot_umaps_batch_correct.py

@@ -38,6 +38,7 @@
 # cmap_choice = "BuPu" # previous default = "viridis"
 bupu = cm.get_cmap('BuPu', 512)
 cmap_choice= ListedColormap(bupu(np.linspace(0.1, 1, 256)))
+dpi_use = 400 # previous default of 300 made the plots look blurry with high cell numbers (200k+)
 
 
 # load metadata
@@ -87,7 +88,7 @@
         os.mkdir(os.path.join(args.fig_dir, mod))
     L.info("plotting modality: %s" % mod)
     plt_df = umaps_df[umaps_df['mod'] == mod].copy()
-    pointsize = 10000 / plt_df.shape[0]
+    pointsize = 100000 / plt_df.shape[0]
     plt_df["method"] = plt_df["method"].astype("category")
     # put none at the top of the list
     if mod != "multimodal":
@@ -113,11 +114,11 @@
         #plt_df = plt_df.sort_values(by="umap_1")
         g = sns.FacetGrid(plt_df, col="method", col_wrap=3, sharex=False, sharey=False)
         g = (g.map(sns.scatterplot, "umap_1", "umap_2", col, s=pointsize, linewidth=0))
-        g.add_legend() 
+        plt.legend(loc = 'center left', bbox_to_anchor = (1, 0.5), markerscale = 20)
         g.savefig(os.path.join(args.fig_dir, mod, "umap_method_" + str(col) + ".png"))
         fig, ax = batch_scatter_two_var(plt_df, "method", col, palette_choice=palette_choice)
         if fig is not None:
-            fig.savefig(os.path.join(args.fig_dir, mod,  "umap_method_facet_" + str(col) + ".png"), dpi=300)
+            fig.savefig(os.path.join(args.fig_dir, mod,  "umap_method_facet_" + str(col) + ".png"), dpi=dpi_use)
         plt.clf()
 
     ncats  = len(plt_df['method'].unique())
@@ -165,7 +166,7 @@
                         cbar_ax = fig.add_axes([0.85, 0.35, 0.025, 0.35])
                         fig.colorbar(im, cbar_ax)
                     fig.suptitle(qc)
-                    plt.savefig(os.path.join(args.fig_dir, mod , "umap_method_" + qc + ".png"), dpi = 300)
+                    plt.savefig(os.path.join(args.fig_dir, mod , "umap_method_" + qc + ".png"), dpi = dpi_use)
                     plt.clf()
                 else:
                     # this is a categorical colored plot
@@ -174,10 +175,10 @@
                         L.info("plotting qc var (categorical) %s"%qc)
                         g = sns.FacetGrid(plt_df, col="method", col_wrap=3, sharex=False, sharey=False)
                         g = (g.map(sns.scatterplot, "umap_1", "umap_2", qc, s=pointsize, linewidth=0))
-                        g.add_legend() 
-                        g.savefig(os.path.join(args.fig_dir, mod, "umap_method_" + qc + ".png"), dpi = 300)
+                        plt.legend(loc = 'center left', bbox_to_anchor = (1, 0.5), markerscale = 20)
+                        g.savefig(os.path.join(args.fig_dir, mod, "umap_method_" + qc + ".png"), dpi = dpi_use)
                         plt.clf()
                     else:
-                        L.info('skipping plot as too many categorys %s' % qc )
+                        L.info('skipping plot as too many categories %s' % qc )
 
 L.info('done')

pyproject.toml

@@ -66,6 +66,11 @@ spatial = [
     "cell2location",
     "tangram-sc"
 ]
+
+multivipatch = [
+    "scvi-tools<=0.20.3",
+    "requests"
+]
 [project.scripts]
 panpipes = "panpipes:entry.main"