add tf ko

app.py

@@ -277,78 +277,85 @@ def get_heatmaps(gene, locus1, locus2, heatmap_range):
     return alpha, beta, delta, pp, acinar, duct, psc
 
 
-def perturb(gene, locus1, locus2, use_max_gene):
-    locus1 = locus1.replace(',', '')
-    locus2 = locus2.replace(',', '')
-    res = {'feat': [], 'log2FC': [], 'var': [], 'a1_idx': [], 'a2_idx': []}
-    for c in celltypes:
-        res[c] = []
-        res[f'{c}_var'] = []
-
-    links_dict = {}
-    for c in celltypes:
-        links_dict[c] = {'chrom1': [], 'chrom1Start': [], 'chrom1End': [], 'chrom2': [], 'chrom2Start': [], 'chrom2End': [], 'score': [], 'strand1': [], 'strand2': []}
-
-    if locus1.startswith('chr'):
-        chrom = locus1.split(':')[0]
-        pos1 = locus1.split(':')[1]
-        a1 = peaks.query(f"chrom == '{chrom}'")['chromStart'].apply(lambda s: abs(s - int(pos1))).idxmin()
-    else:
-        a1 = get_closest_peak_to_gene(locus1, rna, peaks)
-    if locus2.startswith('chr'):
-        chrom = locus2.split(':')[0]
-        pos2 = locus2.split(':')[1]
-        a2 = peaks.query(f"chrom == '{chrom}'")['chromStart'].apply(lambda s: abs(s - int(pos2))).idxmin()
-    else:
-        a2 = get_closest_peak_to_gene(locus2, rna, peaks)
-    
-    print(a1)
-    print(a2)
-
-    a1_closest_gene = get_closest_gene_to_peak(a1, rna, peaks)
-    a2_closest_gene = get_closest_gene_to_peak(a2, rna, peaks)
-    print(a1_closest_gene, a2_closest_gene)
-    
-    guidance_hvf = prior.copy()
-    try:
-        path_to_gene_a1 = nx.shortest_path(prior, source=a1, target=gene)
+def perturb(gene, locus1, locus2, use_max_gene, tf_knockout):
+    if tf_knockout:
+        guidance_hvf = prior.copy()
         try:
-            guidance_hvf.remove_edge(path_to_gene_a1[-2], path_to_gene_a1[-1])
+            guidance_hvf.remove_node(gene)
         except:
             pass
+    else:
+        locus1 = locus1.replace(',', '')
+        locus2 = locus2.replace(',', '')
+        res = {'feat': [], 'log2FC': [], 'var': [], 'a1_idx': [], 'a2_idx': []}
+        for c in celltypes:
+            res[c] = []
+            res[f'{c}_var'] = []
+
+        links_dict = {}
+        for c in celltypes:
+            links_dict[c] = {'chrom1': [], 'chrom1Start': [], 'chrom1End': [], 'chrom2': [], 'chrom2Start': [], 'chrom2End': [], 'score': [], 'strand1': [], 'strand2': []}
+
+        if locus1.startswith('chr'):
+            chrom = locus1.split(':')[0]
+            pos1 = locus1.split(':')[1]
+            a1 = peaks.query(f"chrom == '{chrom}'")['chromStart'].apply(lambda s: abs(s - int(pos1))).idxmin()
+        else:
+            a1 = get_closest_peak_to_gene(locus1, rna, peaks)
+        if locus2.startswith('chr'):
+            chrom = locus2.split(':')[0]
+            pos2 = locus2.split(':')[1]
+            a2 = peaks.query(f"chrom == '{chrom}'")['chromStart'].apply(lambda s: abs(s - int(pos2))).idxmin()
+        else:
+            a2 = get_closest_peak_to_gene(locus2, rna, peaks)
+        
+        print(a1)
+        print(a2)
+
+        a1_closest_gene = get_closest_gene_to_peak(a1, rna, peaks)
+        a2_closest_gene = get_closest_gene_to_peak(a2, rna, peaks)
+        print(a1_closest_gene, a2_closest_gene)
+        
+        guidance_hvf = prior.copy()
         try:
-            guidance_hvf.remove_edge(path_to_gene_a1[-1], path_to_gene_a1[-2])
+            path_to_gene_a1 = nx.shortest_path(prior, source=a1, target=gene)
+            try:
+                guidance_hvf.remove_edge(path_to_gene_a1[-2], path_to_gene_a1[-1])
+            except:
+                pass
+            try:
+                guidance_hvf.remove_edge(path_to_gene_a1[-1], path_to_gene_a1[-2])
+            except:
+                pass
         except:
             pass
-    except:
-        pass
-    try:
-        path_to_gene_a2 = nx.shortest_path(prior, source=a2, target=gene)
         try:
-            guidance_hvf.remove_edge(path_to_gene_a2[-2], path_to_gene_a2[-1])
+            path_to_gene_a2 = nx.shortest_path(prior, source=a2, target=gene)
+            try:
+                guidance_hvf.remove_edge(path_to_gene_a2[-2], path_to_gene_a2[-1])
+            except:
+                pass
+            try:
+                guidance_hvf.remove_edge(path_to_gene_a2[-1], path_to_gene_a2[-2])
+            except:
+                pass
+        except:
+                pass
+        try:
+            guidance_hvf.remove_edge(a1, a2)
+            guidance_hvf.remove_edge(a2, a1)
         except:
             pass
         try:
-            guidance_hvf.remove_edge(path_to_gene_a2[-1], path_to_gene_a2[-2])
+            guidance_hvf.remove_edge(a1, a1_closest_gene)
+            guidance_hvf.remove_edge(a1_closest_gene, a1)
         except:
             pass
-    except:
+        try:
+            guidance_hvf.remove_edge(a2, a2_closest_gene)
+            guidance_hvf.remove_edge(a2_closest_gene, a2)
+        except:
             pass
-    try:
-        guidance_hvf.remove_edge(a1, a2)
-        guidance_hvf.remove_edge(a2, a1)
-    except:
-        pass
-    try:
-        guidance_hvf.remove_edge(a1, a1_closest_gene)
-        guidance_hvf.remove_edge(a1_closest_gene, a1)
-    except:
-        pass
-    try:
-        guidance_hvf.remove_edge(a2, a2_closest_gene)
-        guidance_hvf.remove_edge(a2_closest_gene, a2)
-    except:
-        pass
 
     perterbed_rna_recon = glue.decode_data("rna", "rna", rna, guidance_hvf)
     ism = np.log2((perterbed_rna_recon + eps) / (rna_recon + eps))
@@ -364,7 +371,7 @@ def perturb(gene, locus1, locus2, use_max_gene):
     rna.obs['old_mean'] = rna_recon[:, gene_idx]
     rna.obs['new_mean'] = perterbed_rna_recon[:, gene_idx] 
     # compute new count based on log2FC
-    rna.obs['new_count'] = rna.layers['counts'][:, gene_idx] * 2 ** rna.obs['log2FC']
+    rna.obs['new_count'] = (rna.layers['counts'][:, gene_idx] + eps) * 2 ** rna.obs['log2FC'] - eps
     
     fig = sc.pl.umap(rna, 
                      color=['celltype', 'log2FC'], 
@@ -422,12 +429,14 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
     with gr.Row():
         with gr.Column():
             in_locus = gr.Textbox(label="Target Gene", elem_id='in-locus', scale=1)
-            max_gene_checkbox = gr.Checkbox(label="Max Gene", elem_id='max-gene-checkbox', scale=1, checked=False)
+            with gr.Row():
+                max_gene_checkbox = gr.Checkbox(label="Max Gene", elem_id='max-gene-checkbox', scale=1, checked=False)
+                tf_ko = gr.Checkbox(label="TF KO", elem_id='tf-ko', scale=1, checked=False)
             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)
             heatmap_x = gr.Textbox(label="Heatmap X", elem_id='heatmap-x', scale=1, visible=False)
             heatmap_y = gr.Textbox(label="Heatmap Y", elem_id='heatmap-y', scale=1, visible=False)
-            heatmap_size = gr.Slider(label="Heatmap Range", info="kb range aroung input gene locus to expand", minimum=50, maximum=5000, value=500)
+            heatmap_size = gr.Slider(label="Heatmap Range", info="kb range aroung input gene locus to expand", minimum=50, maximum=5000, value=250)
     heatmap_button = gr.Button(value="Generate Heatmaps", elem_id='heatmap-button', scale=1)
     with gr.Row():
         out_heatmaps = []
@@ -443,7 +452,7 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
         out_violin_fc = gr.Image(elem_id='out-violin-fc', scale=1)
         #out_plot = gr.Plot(elem_id='out-plot', scale=1)
 
-    inputs = [in_locus, anchor1, anchor2, max_gene_checkbox]
+    inputs = [in_locus, anchor1, anchor2, max_gene_checkbox, tf_ko]
     outputs = [out_img, out_violin, out_violin_new, out_violin_fc]
 
     gr.Examples(examples=[['INS', 'chr11:2,289,895', 'chr11:2,298,840', False],
@@ -456,6 +465,7 @@ with gr.Blocks(theme='WeixuanYuan/Soft_dark') as demo:
                           ['LOXL4', 'chr10:100186215', 'chr10:99913493', False],
                           ['KRT19', 'chr17:22,220,637', 'chr17:39,591,813', False],
                           ['LPP', 'chr3:188,097,749', 'chr3:197,916,262', False],
+                          ['MAFB', 'chr20:39431654', 'chr20:39368271', True],
                           ['CEL', 'chr9:135,937,365', 'chr9:135,973,107', False]],
                 inputs=inputs,
                 outputs=outputs,