# Dash app to visualize scRNA-seq data quality control metrics from scanpy objects # Shoutout to Coding-with-Adam for the initial template of the project: # https://github.com/Coding-with-Adam/Dash-by-Plotly/blob/master/Dash%20Components/Graph/dash-graph.py import dash from dash import dcc, html, Output, Input import plotly.express as px import dash_callback_chain import yaml import polars as pl import os pl.enable_string_cache(False) # Set custom resolution for plots: config_fig = { 'toImageButtonOptions': { 'format': 'svg', 'filename': 'custom_image', 'height': 600, 'width': 700, 'scale': 1, } } from adlfs import AzureBlobFileSystem mountpount=os.environ['AZURE_MOUNT_POINT'], AZURE_STORAGE_ACCESS_KEY=os.getenv('AZURE_STORAGE_ACCESS_KEY') AZURE_STORAGE_ACCOUNT=os.getenv('AZURE_STORAGE_ACCOUNT') # Load in config file config_path = "./data/config.yaml" # Add the read-in data from the yaml file def read_config(filename): with open(filename, 'r') as yaml_file: config = yaml.safe_load(yaml_file) return config config = read_config(config_path) path_parquet = config.get("path_parquet") col_batch = config.get("col_batch") col_features = config.get("col_features") col_counts = config.get("col_counts") col_mt = config.get("col_mt") filepath = f"az://{path_parquet}" storage_options={'account_name': AZURE_STORAGE_ACCOUNT, 'account_key': AZURE_STORAGE_ACCESS_KEY,'anon': False} #azfs = AzureBlobFileSystem(**storage_options ) # Load in multiple dataframes df = pl.read_parquet(filepath, storage_options=storage_options) # Setup the app external_stylesheets = ['https://codepen.io/chriddyp/pen/bWLwgP.css'] app = dash.Dash(__name__, external_stylesheets=external_stylesheets) #, requests_pathname_prefix='/dashboard1/' #df = pl.read_parquet(filepath,storage_options=storage_options) #df = pl.DataFrame() #abfs = AzureBlobFileSystem(account_name=accountname,account_key=accountkey) #df = df.rename({"__index_level_0__": "Unnamed: 0"}) #df1 = pl.read_parquet(filepath, storage_options=storage_options) #df2 = pl.read_parquet(f"az://data10xflex/{dataset_chosen}.parquet", storage_options=storage_options) #tab0_content = html.Div([ # html.Label("Dataset chosen"), # dcc.Dropdown(id='dpdn1', value="corg/10xflexcorg_umap_clusres", multi=False, # options=["corg/10xflexcorg_umap_clusres","d1011/10xflexd1011_umap_clusres"]) #]) #@app.callback( # Input(component_id='dpdn1', component_property='value') #) #def update_filepath(dpdn1): # global df # if str(f"az://data10xflex/{dpdn1}.parquet") != str(filepath): # print("not identical filepath, chosing other") # df2 = pl.read_parquet(f"az://data10xflex/{dpdn1}.parquet", storage_options=storage_options) # df = df2 # return #df = pl.read_parquet(filepath, storage_options=storage_options) min_value = df[col_features].min() max_value = df[col_features].max() min_value_2 = df[col_counts].min() min_value_2 = round(min_value_2) max_value_2 = df[col_counts].max() max_value_2 = round(max_value_2) min_value_3 = df[col_mt].min() min_value_3 = round(min_value_3, 1) max_value_3 = df[col_mt].max() max_value_3 = round(max_value_3, 1) # Loads in the conditions specified in the yaml file # Note: Future version perhaps all values from a column in the dataframe of the parquet file # Note 2: This could also be a tsv of the categories and own specified colors #conditions = df[col_batch].unique().to_list() # Create the first tab content # Add Sliders for three QC params: N genes by counts, total amount of reads and pct MT reads tab1_content = html.Div([ html.Label("Column chosen"), dcc.Dropdown(id='dpdn2', value="batch", multi=False, options=df.columns), html.Label("N Genes by Counts"), dcc.RangeSlider( id='range-slider-1', step=250, value=[min_value, max_value], marks={i: str(i) for i in range(min_value, max_value + 1, 250)}, ), dcc.Input(id='min-slider-1', type='number', value=min_value, debounce=True), dcc.Input(id='max-slider-1', type='number', value=max_value, debounce=True), html.Label("Total Counts"), dcc.RangeSlider( id='range-slider-2', step=7500, value=[min_value_2, max_value_2], marks={i: str(i) for i in range(min_value_2, max_value_2 + 1, 7500)}, ), dcc.Input(id='min-slider-2', type='number', value=min_value_2, debounce=True), dcc.Input(id='max-slider-2', type='number', value=max_value_2, debounce=True), html.Label("Percent Mitochondrial Genes"), dcc.RangeSlider( id='range-slider-3', step=5, min=0, max=100, value=[min_value_3, max_value_3], ), dcc.Input(id='min-slider-3', type='number', value=min_value_3, debounce=True), dcc.Input(id='max-slider-3', type='number', value=max_value_3, debounce=True), html.Div([ dcc.Graph(id='pie-graph', figure={}, className='four columns',config=config_fig), dcc.Graph(id='my-graph', figure={}, clickData=None, hoverData=None, className='four columns',config=config_fig ), dcc.Graph(id='scatter-plot', figure={}, className='four columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-2', figure={}, className='four columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-3', figure={}, className='four columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-4', figure={}, className='four columns',config=config_fig) ]), ]) # Create the second tab content with scatter-plot-5 and scatter-plot-6 tab2_content = html.Div([ html.Div([ html.Label("S-cycle genes"), dcc.Dropdown(id='dpdn3', value="MCM5", multi=False, options=[ "MCM5", "PCNA", "TYMS", "FEN1", "MCM2", "MCM4", "RRM1", "UNG", "GINS2", "MCM6", "CDCA7", "DTL", "PRIM1", "UHRF1", "MLF1IP", "HELLS", "RFC2", "RPA2", "NASP", "RAD51AP1", "GMNN", "WDR76", "SLBP", "CCNE2", "UBR7", "POLD3", "MSH2", "ATAD2", "RAD51", "RRM2", "CDC45", "CDC6", "EXO1", "TIPIN", "DSCC1", "BLM", "CASP8AP2", "USP1", "CLSPN", "POLA1", "CHAF1B", "BRIP1", "E2F8" ]), html.Label("G2M-cycle genes"), dcc.Dropdown(id='dpdn4', value="TOP2A", multi=False, options=[ 'HMGB2', 'CDK1', 'NUSAP1', 'UBE2C', 'BIRC5', 'TPX2', 'TOP2A', 'NDC80', 'CKS2', 'NUF2', 'CKS1B', 'MKI67', 'TMPO', 'CENPF', 'TACC3', 'FAM64A', 'SMC4', 'CCNB2', 'CKAP2L', 'CKAP2', 'AURKB', 'BUB1', 'KIF11', 'ANP32E', 'TUBB4B', 'GTSE1', 'KIF20B', 'HJURP', 'CDCA3', 'HN1', 'CDC20', 'TTK', 'CDC25C', 'KIF2C', 'RANGAP1', 'NCAPD2', 'DLGAP5', 'CDCA2', 'CDCA8', 'ECT2', 'KIF23', 'HMMR', 'AURKA', 'PSRC1', 'ANLN', 'LBR', 'CKAP5', 'CENPE', 'CTCF', 'NEK2', 'G2E3', 'GAS2L3', 'CBX5', 'CENPA' ]), ]), html.Div([ dcc.Graph(id='scatter-plot-5', figure={}, className='three columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-6', figure={}, className='three columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-7', figure={}, className='three columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-8', figure={}, className='three columns',config=config_fig) ]), ]) # Create the second tab content with scatter-plot-5 and scatter-plot-6 tab3_content = html.Div([ html.Div([ html.Label("UMAP condition 1"), dcc.Dropdown(id='dpdn5', value="batch", multi=False, options=df.columns), html.Label("UMAP condition 2"), dcc.Dropdown(id='dpdn6', value="n_genes_by_counts", multi=False, options=df.columns), html.Div([ dcc.Graph(id='scatter-plot-9', figure={}, className='four columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-10', figure={}, className='four columns',config=config_fig) ]), html.Div([ dcc.Graph(id='scatter-plot-11', figure={}, className='four columns',config=config_fig) ]), html.Div([ dcc.Graph(id='my-graph2', figure={}, clickData=None, hoverData=None, className='four columns',config=config_fig ) ]), ]), ]) # html.Div([ # dcc.Graph(id='scatter-plot-12', figure={}, className='four columns',config=config_fig) # ]), tab4_content = html.Div([ html.Div([ html.Label("Multi gene"), dcc.Dropdown(id='dpdn7', value=["PAX6","TP63","S100A9"], multi=True, options=df.columns), ]), html.Div([ dcc.Graph(id='scatter-plot-12', figure={}, className='four columns',config=config_fig) ]), ]) # Define the tabs layout app.layout = html.Div([ dcc.Tabs(id='tabs', style= {'width': 600, 'font-size': '100%', 'height': 50}, value='tab1',children=[ #dcc.Tab(label='Dataset', value='tab0', children=tab0_content), dcc.Tab(label='QC', value='tab1', children=tab1_content), dcc.Tab(label='Cell cycle', value='tab2', children=tab2_content), dcc.Tab(label='Custom', value='tab3', children=tab3_content), dcc.Tab(label='Multi dot', value='tab4', children=tab4_content), ]), ]) # Define the circular callback @app.callback( Output("min-slider-1", "value"), Output("max-slider-1", "value"), Output("min-slider-2", "value"), Output("max-slider-2", "value"), Output("min-slider-3", "value"), Output("max-slider-3", "value"), Input("min-slider-1", "value"), Input("max-slider-1", "value"), Input("min-slider-2", "value"), Input("max-slider-2", "value"), Input("min-slider-3", "value"), Input("max-slider-3", "value"), ) def circular_callback(min_1, max_1, min_2, max_2, min_3, max_3): return min_1, max_1, min_2, max_2, min_3, max_3 @app.callback( Output('range-slider-1', 'value'), Output('range-slider-2', 'value'), Output('range-slider-3', 'value'), Input('min-slider-1', 'value'), Input('max-slider-1', 'value'), Input('min-slider-2', 'value'), Input('max-slider-2', 'value'), Input('min-slider-3', 'value'), Input('max-slider-3', 'value'), ) def update_slider_values(min_1, max_1, min_2, max_2, min_3, max_3): return [min_1, max_1], [min_2, max_2], [min_3, max_3] @app.callback( Output(component_id='my-graph', component_property='figure'), Output(component_id='pie-graph', component_property='figure'), Output(component_id='scatter-plot', component_property='figure'), Output(component_id='scatter-plot-2', component_property='figure'), Output(component_id='scatter-plot-3', component_property='figure'), Output(component_id='scatter-plot-4', component_property='figure'), # Add this new scatter plot Output(component_id='scatter-plot-5', component_property='figure'), Output(component_id='scatter-plot-6', component_property='figure'), Output(component_id='scatter-plot-7', component_property='figure'), Output(component_id='scatter-plot-8', component_property='figure'), Output(component_id='scatter-plot-9', component_property='figure'), Output(component_id='scatter-plot-10', component_property='figure'), Output(component_id='scatter-plot-11', component_property='figure'), Output(component_id='scatter-plot-12', component_property='figure'), Output(component_id='my-graph2', component_property='figure'), Input(component_id='dpdn2', component_property='value'), Input(component_id='dpdn3', component_property='value'), Input(component_id='dpdn4', component_property='value'), Input(component_id='dpdn5', component_property='value'), Input(component_id='dpdn6', component_property='value'), Input(component_id='dpdn7', component_property='value'), Input(component_id='range-slider-1', component_property='value'), Input(component_id='range-slider-2', component_property='value'), Input(component_id='range-slider-3', component_property='value') ) def update_graph_and_pie_chart(col_chosen, s_chosen, g2m_chosen, condition1_chosen, condition2_chosen, condition3_chosen, range_value_1, range_value_2, range_value_3): #batch_chosen, batch_chosen = df[col_chosen].unique().to_list() dff = df.filter( (pl.col(col_chosen).cast(str).is_in(batch_chosen)) & (pl.col(col_features) >= range_value_1[0]) & (pl.col(col_features) <= range_value_1[1]) & (pl.col(col_counts) >= range_value_2[0]) & (pl.col(col_counts) <= range_value_2[1]) & (pl.col(col_mt) >= range_value_3[0]) & (pl.col(col_mt) <= range_value_3[1]) ) #Drop categories that are not in the filtered data dff = dff.with_columns(dff[col_chosen].cast(pl.Categorical)) dff = dff.sort(col_chosen) # Plot figures fig_violin = px.violin(data_frame=dff, x=col_chosen, y=col_features, box=True, points="all", color=col_chosen, hover_name=col_chosen,template="seaborn") # Cache commonly used subexpressions total_count = pl.lit(len(dff)) category_counts = dff.group_by(col_chosen).agg(pl.col(col_chosen).count().alias("count")) category_counts = category_counts.with_columns(((pl.col("count") / total_count * 100).round(decimals=2)).alias("normalized_count")) # Sort the dataframe #category_counts = category_counts.sort(col_chosen) does not work check if the names are different ... # Display the result total_cells = total_count # Calculate total number of cells pie_title = f'Percentage of Total Cells: {total_cells}' # Include total cells in the title # Calculate the mean expression # Melt wide format DataFrame into long format # Specify batch column as string type and gene columns as float type list_conds = condition3_chosen list_conds += [col_chosen] dff_pre = dff.select(list_conds) # Melt wide format DataFrame into long format dff_long = dff_pre.melt(id_vars=col_chosen, variable_name="Gene", value_name="Mean expression") # Calculate the mean expression levels for each gene in each region expression_means = dff_long.lazy().group_by([col_chosen, "Gene"]).agg(pl.mean("Mean expression")).collect() # Calculate the percentage total expressed dff_long1 = dff_pre.melt(id_vars=col_chosen, variable_name="Gene")#.group_by(pl.all()).agg(pl.len()) count = 1 dff_long2 = dff_long1.with_columns(pl.lit(count).alias("len")) dff_long3 = dff_long2.filter(pl.col("value") > 0).group_by([col_chosen, "Gene"]).agg(pl.sum("len").alias("len")) dff_long4 = dff_long2.group_by([col_chosen, "Gene"]).agg(pl.sum("len").alias("total")) dff_5 = dff_long4.join(dff_long3, on=[col_chosen,"Gene"], how="outer") result = dff_5.select([ pl.when((pl.col('len').is_not_null()) & (pl.col('total').is_not_null())) .then(pl.col('len') / pl.col('total')*100) .otherwise(None).alias("%"), ]) result = result.with_columns(pl.col("%").fill_null(100)) dff_5[["percentage"]] = result[["%"]] dff_5 = dff_5.select(pl.col(col_chosen,"Gene","percentage")) # Final part to join the percentage expressed and mean expression levels # TO DO expression_means = expression_means.join(dff_5, on=[col_chosen,"Gene"], how="inner") # Order the dataframe on ascending categories expression_means = expression_means.sort(col_chosen, descending=True) #expression_means = expression_means.select(["batch", "Gene", "Expression"] + condition3_chosen) category_counts = category_counts.sort(col_chosen) fig_pie = px.pie(category_counts, values="normalized_count", names=col_chosen, labels=col_chosen, hole=.3, title=pie_title, template="seaborn") #labels = category_counts[col_chosen].to_list() #values = category_counts["normalized_count"].to_list() # Create the scatter plots fig_scatter = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=col_chosen, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_2 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=col_mt, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_3 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=col_features, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_4 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=col_counts, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_5 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=s_chosen, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch', title="S-cycle gene:",template="seaborn") fig_scatter_6 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=g2m_chosen, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch', title="G2M-cycle gene:",template="seaborn") fig_scatter_7 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color="S_score", labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch', title="S score:",template="seaborn") fig_scatter_8 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color="G2M_score", labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch', title="G2M score:",template="seaborn") # Sort values of custom in-between dff = dff.sort(condition1_chosen) fig_scatter_9 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=condition1_chosen, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_10 = px.scatter(data_frame=dff, x='X_umap-0', y='X_umap-1', color=condition2_chosen, labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_11 = px.scatter(data_frame=dff, x=condition1_chosen, y=condition2_chosen, color=condition1_chosen, #labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name='batch',template="seaborn") fig_scatter_12 = px.scatter(data_frame=expression_means, x="Gene", y=col_chosen, color="Mean expression", size="percentage", size_max = 20, #labels={'X_umap-0': 'umap1' , 'X_umap-1': 'umap2'}, hover_name=col_chosen,template="seaborn") fig_violin2 = px.violin(data_frame=dff, x=condition1_chosen, y=condition2_chosen, box=True, points="all", color=condition1_chosen, hover_name=condition1_chosen,template="seaborn") return fig_violin, fig_pie, fig_scatter, fig_scatter_2, fig_scatter_3, fig_scatter_4, fig_scatter_5, fig_scatter_6, fig_scatter_7, fig_scatter_8, fig_scatter_9, fig_scatter_10, fig_scatter_11, fig_scatter_12, fig_violin2 # Set http://localhost:5000/ in web browser # Now create your regular FASTAPI application if __name__ == '__main__': app.run_server(debug=False, use_reloader=False, host='0.0.0.0', port=5000) #