more cleanup and prettyify

app.py

@@ -9,8 +9,6 @@ import shutil
 import numpy as np
 import matplotlib.pyplot as plt
 
-from itertools import chain
-
 from huggingface_hub import hf_hub_download
 
 
@@ -58,10 +56,7 @@ color_map['Other'] = 'gray'
 
 hic = ad.read_h5ad(hic_file)
 hic.var["highly_variable"] = hic.var[f"{prior_name}_highly_variable"]
-print(hic)
-print('Loading prior knowledge graph...')
 prior = nx.read_graphml(graph_file)
-#scglue.models.configure_dataset(hic, "NB", use_highly_variable=True)
 glue = scglue.models.load_model(model_file)
 
 
@@ -76,18 +71,14 @@ for e, attr in dict(prior.edges).items():
             genes.append(gene_name)
     elif attr["type"] == 'hic':
         loops.append(e)
-print(len(genes))
 rna.var["highly_variable"] = rna.var["highly_variable"] & rna.var["in_hic"]
 genes = rna.var.query(f"highly_variable").index
-print(len(genes))
 gene_idx_map = {}
 for i, gene in enumerate(genes):
     gene_idx_map[gene] = i
-peaks = hic.var.query("highly_variable")
+peaks = hic.var.query("highly_variable").copy()
 
 rna_recon = glue.decode_data("rna", "rna", rna, prior)
-#hic_recon = glue.decode_data("hic", "hic", hic, prior)
-print(rna_recon.shape)
 
 
 def get_closest_peak_to_gene(gene_name, rna, peaks):
@@ -118,7 +109,7 @@ def get_chromosome_from_filename(filename):
 def perturb(gene, locus1, locus2):
     locus1 = locus1.replace(',', '')
     locus2 = locus2.replace(',', '')
-    res = {'feat': [], 'ism': [], 'var': [], 'a1_idx': [], 'a2_idx': []}
+    res = {'feat': [], 'log2FC': [], 'var': [], 'a1_idx': [], 'a2_idx': []}
     for c in celltypes:
         res[c] = []
         res[f'{c}_var'] = []
@@ -181,12 +172,12 @@ def perturb(gene, locus1, locus2):
     print('Max gene:', max_gene)
     # get integer index of gene
     gene_idx = gene_idx_map[gene]
-    rna.obs['ism'] = ism[:, gene_idx]
+    rna.obs['log2FC'] = ism[:, gene_idx]
     
     fig = sc.pl.umap(rna, 
-                     color=["celltype", "ism"], 
+                     color=['celltype', 'log2FC'], 
                      color_map='Spectral', 
-                     wspace=0.15, 
+                     wspace=0.05, 
                      legend_loc='on data', 
                      legend_fontoutline=2,
                      frameon=False, 
@@ -204,24 +195,14 @@ def perturb(gene, locus1, locus2):
 
     return image_from_plot
 
-    # if max_gene != gene:
-    #     rna.obs['ism_max'] = ism[:, max_gene_idx]
-    #     fig = sc.pl.umap(rna, color=["celltype", "ism_max"], color_map='Spectral', wspace=0.35, return_fig=True)
-    #     filename = os.path.join(plot_dir, f'{gene}/islet_glue_rna_umap_perturb_{a1}_{a2}_{max_gene}.png')
-    #     fig.suptitle(f'{max_gene} after removing ' + a1 + '-' + a2)
-    #     fig.tight_layout()
-    #     fig.savefig(filename)
-    #     #fig.savefig(filename.replace('.png', '.pdf'))
-    #     plt.close()
-
 
 
 with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
     with gr.Row():
         with gr.Column():
-            in_locus = gr.Textbox(label="Gene", elem_id='in-locus', scale=1)
-            anchor1 = gr.Textbox(label="Locus 1", elem_id='anchor1', scale=1)
-            anchor2 = gr.Textbox(label="Locus 2", elem_id='anchor2', scale=1)
+            in_locus = gr.Textbox(label="Target Gene", elem_id='in-locus', scale=1)
+            anchor1 = gr.Textbox(label="Locus 1 (gene or genomic coords)", elem_id='anchor1', scale=1)
+            anchor2 = gr.Textbox(label="Locus 2 (gene or genomic coords)", elem_id='anchor2', scale=1)
     with gr.Row():
         run_button = gr.Button(label="Run", elem_id='run-button', scale=1)
     with gr.Row():
@@ -230,7 +211,9 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
 
     gr.Examples(examples=[['INS', 'chr11:2,289,895', 'chr11:2,298,840'],
                           ['INS', 'INS', 'IGF1'],
+                          ['TSPAN1', 'TSPAN1', 'PIK3R3'],
                           ['GCG', 'GCG', 'FAP'],
+                          ['KRT19', 'chr17:22,220,637', 'chr17:39,591,813'],
                           ['LPP', 'chr3:188,097,749', 'chr3:197,916,262'],
                           ['CEL', 'chr9:135,937,365', 'chr9:135,973,107']],
                 inputs=[in_locus, anchor1, anchor2],