1-samples_selection.py

from Packages import *
st.set_page_config(page_title="NIRS Utils", page_icon=":goat:", layout="wide")
from Modules import *

# empty temp figures
repertoire_a_vider = Path('Report/figures')
if os.path.exists(repertoire_a_vider):
    for fichier in os.listdir(repertoire_a_vider):
        chemin_fichier = os.path.join(repertoire_a_vider, fichier)
        if os.path.isfile(chemin_fichier) or os.path.islink(chemin_fichier):
            os.unlink(chemin_fichier)
        elif os.path.isdir(chemin_fichier):
            shutil.rmtree(chemin_fichier)
# HTML pour le bandeau "CEFE - CNRS"
add_header()
#load specific model page css
local_css(css_file / "style_model.css")

#define some variables
tcr=pd.DataFrame()
sam=pd.DataFrame()
sam1=pd.DataFrame()
dim_red_methods=['', 'PCA','UMAP', 'NMF']  # List of dimensionality reduction algos
cluster_methods = ['', 'Kmeans','HDBSCAN', 'AP'] # List of clustering algos
selec_strategy = ['center','random']

# check session state and define default values if simple interface to automate processing
st.session_state["interface"] = st.session_state.get('interface')
if st.session_state["interface"] == 'simple':
    st.write(':red[Automated Simple Interface]')
    hide_pages("Predictions")
    if 37 not in st.session_state:
        default_reduction_option = 1
    else:
        default_reduction_option = dim_red_methods.index(st.session_state.get(37))
    if 38 not in st.session_state:
        default_clustering_option = 1
    else:
        default_clustering_option = cluster_methods.index(st.session_state.get(38))
    if 102 not in st.session_state:
        default_sample_selection_option = 1
    else:
        default_sample_selection_option = selec_strategy.index(st.session_state.get(102))

if st.session_state["interface"] == 'advanced':
    default_reduction_option = 0
    default_clustering_option = 0
    default_sample_selection_option = 0

################################### I - Data Loading and Visualization ########################################
st.header("I - Spectral Data Visualization", divider='blue')
col2, col1 = st.columns([3, 1])


## Preallocation of data structure
spectra = pd.DataFrame
meta_data = pd.DataFrame
selected_samples = pd.DataFrame
non_clustered = None
colnames = []
rownames = []
l1 = []
# loader for datafile
data_file = col1.file_uploader("Load NIRS Data", type=["csv","dx"], help=" :mushroom: select a csv matrix with samples as rows and lambdas as columns", key=5)

if data_file:
    # Retrieve the extension of the file
    test = data_file.name[data_file.name.find('.'):]

    ## Load .csv file
    if test== '.csv':
        with col1:
            # Select list for CSV delimiter
            psep = st.radio("Select csv separator - _detected_: " + str(find_delimiter('data/'+data_file.name)), options=[";", ","], index=[";", ","].index(str(find_delimiter('data/'+data_file.name))), key=9)
                # Select list for CSV header True / False
            phdr = st.radio("indexes column in csv? - _detected_: " + str(find_col_index('data/'+data_file.name)), options=["no", "yes"], index=["no", "yes"].index(str(find_col_index('data/'+data_file.name))), key=31)
            if phdr == 'yes':
                col = 0
            else:
                col = False
            imp = pd.read_csv(data_file, sep=psep, index_col=col)
            # spectra = col_cat(imp)[0]
            # meta_data = col_cat(imp)[1]
            spectra, md_df_st_ = col_cat(imp)
            meta_data = md_df_st_
            st.success("The data have been loaded successfully", icon="✅")

    ## Load .dx file
    elif test == '.dx':
        # Create a temporary file to save the uploaded file
        with NamedTemporaryFile(delete=False, suffix=".dx") as tmp:
            tmp.write(data_file.read())
            tmp_path = tmp.name
            with col1:
                _, spectra, meta_data, md_df_st_ = read_dx(file = tmp_path)
                st.success("The data have been loaded successfully", icon="✅")
        os.unlink(tmp_path)


    
## Visualize spectra
if not spectra.empty:
    # retrieve columns name and rows name of spectra
    colnames = list(spectra.columns)
    rownames = [str(i) for i in list(spectra.index)]
    spectra.index = rownames

    with col2:
        if test =='.dx':
            if meta_data.loc[:,'xunits'][0] == '1/cm':
                lab = 'Wavenumber (1/cm)'
            else:
                lab = 'Wavelength (nm)'
            fig = plot_spectra(spectra, xunits = lab, yunits = meta_data.loc[:,'yunits'][0])
        else:
            fig = plot_spectra(spectra, xunits = 'Wavelength/Wavenumber', yunits = 'Signal intensity')

        st.pyplot(fig)
        fig.savefig("./Report/figures/Spectra_Plot.png")


############################## Exploratory data analysis ###############################
st.header("II - Exploratory Data Analysis-Multivariable Data Analysis", divider='blue')
scores, loadings, pc = st.columns([2, 3, 0.5])
influence, hotelling, qexp = st.columns([2, 2, 1])
st.header('III - Selected samples for chemical analysis', divider='blue')

dr_model = None # dimensionality reduction model
cl_model = None # clustering model

###### 1- Dimensionality reduction ######
t = pd.DataFrame # scores
p = pd.DataFrame # loadings
labels = []
if not spectra.empty:
    dim_red_method = pc.selectbox("Dimensionality reduction techniques: ", options = dim_red_methods, index = default_reduction_option, key = 37)
    clus_method = pc.selectbox("Clustering techniques: ", options = cluster_methods, index = default_clustering_option, key = 38)
    xc = standardize(spectra, center=True, scale=False)


    if dim_red_method == dim_red_methods[1]:
        dr_model = LinearPCA(xc, Ncomp=8)

    elif dim_red_method == dim_red_methods[2]:
        if not meta_data.empty:
            filter = md_df_st_.columns
            filter = filter.insert(0, 'Nothing')
            col = pc.selectbox('Supervised UMAP by:', options= filter, key=108)
            if col == 'Nothing':
                supervised = None
            else:
                supervised = md_df_st_[col]
        else:
            supervised = None
        dr_model = Umap(numerical_data = MinMaxScale(spectra), cat_data = supervised)

    elif dim_red_method == dim_red_methods[3]:
        dr_model = Nmf(spectra, Ncomp= 3)

    if dr_model:
        axis1 = pc.selectbox("x-axis", options = dr_model.scores_.columns, index=0)
        axis2 = pc.selectbox("y-axis", options = dr_model.scores_.columns, index=1)
        axis3 = pc.selectbox("z-axis", options = dr_model.scores_.columns, index=2)

        t = pd.concat([dr_model.scores_.loc[:,axis1], dr_model.scores_.loc[:,axis2], dr_model.scores_.loc[:,axis3]], axis = 1)



###### II - clustering #######
if not t.empty:
    tcr = standardize(t)
        # Clustering
    # 1- K-MEANS Clustering
    if clus_method == cluster_methods[1]:
        cl_model = Sk_Kmeans(tcr, max_clusters = 25)
        ncluster = scores.number_input(min_value=2, max_value=25, value=cl_model.suggested_n_clusters_, label = 'Select the desired number of clusters')
        fig2 = px.scatter(cl_model.inertia_.T, y = 'inertia')
        scores.write(f"Suggested n_clusters : {cl_model.suggested_n_clusters_}")
        scores.plotly_chart(fig2,use_container_width=True)
        img = pio.to_image(fig2, format="png")
        with open("./Report/figures/Elbow.png", "wb") as f:
                f.write(img)    
        data, labels, clu_centers = cl_model.fit_optimal(nclusters = ncluster)

    # 2- HDBSCAN clustering
    elif clus_method == cluster_methods[2]:
        optimized_hdbscan = Hdbscan(np.array(tcr))
        # all_labels, hdbscan_score, clu_centers = optimized_hdbscan.HDBSCAN_scores_
        all_labels, clu_centers = optimized_hdbscan.HDBSCAN_scores_
        labels = [f'cluster#{i+1}' if i !=-1 else 'Non clustered' for i in all_labels]

    # 3- Affinity propagation
    elif clus_method == cluster_methods[3]:
        cl_model = AP(X = tcr)
        data, labels, clu_centers = cl_model.fit_optimal_
    
    if clus_method == cluster_methods[2]:
        #clustered = np.where(np.array(labels) != 'Non clustered')[0]
        clustered = np.arange(tcr.shape[0])
        non_clustered = np.where(np.array(labels) == 'Non clustered')[0]

    else:
        clustered = np.arange(tcr.shape[0])
        non_clustered = None
    
    new_tcr = tcr.iloc[clustered,:]    
    

#################################################### III - Samples selection using the reduced data preentation ######
samples_df_chem = pd.DataFrame
selected_samples = []
selected_samples_idx = []


if labels:
    if clus_method:
        selection = scores.radio('Select samples selection strategy:',
                                    options = selec_strategy, index = default_sample_selection_option, key=102)
    # Strategy 0
    if selection == selec_strategy[0]:
        # list samples at clusters centers - Use sklearn.metrics.pairwise_distances_argmin if you want more than 1 sample per cluster
        closest, _ = pairwise_distances_argmin_min(clu_centers, new_tcr)
        selected_samples_idx = np.array(new_tcr.index)[list(closest)]
        selected_samples_idx = selected_samples_idx.tolist()
        
    #### Strategy 1
    elif selection == selec_strategy[1]:
        selection_number = scores.number_input('How many samples per cluster?',
                                                min_value = 1, step=1, value = 3)
        s = np.array(labels)[np.where(np.array(labels) !='Non clustered')[0]]
        for i in np.unique(s):
            C = np.where(np.array(labels) == i)[0]
            if C.shape[0] >= selection_number:
                # scores.write(list(tcr.index)[labels== i])
                km2 = KMeans(n_clusters = selection_number)
                km2.fit(tcr.iloc[C,:])
                clos, _ = pairwise_distances_argmin_min(km2.cluster_centers_, tcr.iloc[C,:])
                selected_samples_idx.extend(tcr.iloc[C,:].iloc[list(clos)].index)
            else:
                selected_samples_idx.extend(new_tcr.iloc[C,:].index.to_list())
            # list indexes of selected samples for colored plot    

    if selected_samples_idx:
        if meta_data.empty:
            sam1 = pd.DataFrame({'name': spectra.index[clustered][selected_samples_idx],
                                'cluster':np.array(labels)[clustered][selected_samples_idx]},
                                index = selected_samples_idx)
        else:
            sam1 = meta_data.iloc[clustered,:].iloc[selected_samples_idx,:]
            sam1.insert(loc=0, column='index', value=selected_samples_idx)
            sam1.insert(loc=1, column='cluster', value=np.array(labels)[selected_samples_idx])
        sam1.index = np.arange(len(selected_samples_idx))+1
        st.write(f' - The total number of samples: {tcr.shape[0]}.\n- The number of selected samples for chemical analysis: {sam1.shape[0]} - {round(sam1.shape[0]/tcr.shape[0]*100, 1)}%.')
        sam = sam1
        if clus_method == cluster_methods[2]:
            unclus = st.checkbox("Include non clustered samples (for HDBSCAN clustering)", value=True)

        if clus_method == cluster_methods[2]:
            if selected_samples_idx:
                if unclus:
                    if meta_data.empty:
                        sam2 = pd.DataFrame({'name': spectra.index[non_clustered],
                                            'cluster':['Non clustered']*len(spectra.index[non_clustered])},
                                            index = spectra.index[non_clustered])
                    else :
                        sam2 = meta_data.iloc[non_clustered,:]
                        sam2.insert(loc=0, column='index', value= spectra.index[non_clustered])
                        sam2.insert(loc=1, column='cluster', value=['Non clustered']*len(spectra.index[non_clustered]))
                    
                    sam = pd.concat([sam1, sam2], axis = 0)
                    sam.index = np.arange(sam.shape[0])+1
                    st.write(f' The number of Non-clustered samples is {sam2.shape[0]} samples. Total selected samples: {sam1.shape[0] + sam2.shape[0]} - {round((sam1.shape[0] + sam2.shape[0]) / tcr.shape[0] * 100, 1)}%.')
        else:
            sam = sam1
        st.write(sam)
################################      Plots visualization          ############################################
 

    ## Scores
if not t.empty:
    with scores:
        fig1, ((ax1, ax2),(ax3,ax4)) = plt.subplots(2,2)
        st.write('Scores plot')
        # scores plot with clustering
        if list(labels) and meta_data.empty:
            fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3, color = labels)
            sns.scatterplot(data = tcr, x = axis1, y =axis2 , hue = labels, ax = ax1)
            

    
        # scores plot with metadata
        elif len(list(labels)) == 0 and not meta_data.empty:
            filter = md_df_st_.columns
            col = st.selectbox('Color by:', options= filter)
            if col == 0:
                fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3)
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , ax = ax1)
                sns.scatterplot(data = tcr, x = axis2, y =axis3 , ax = ax2)
                sns.scatterplot(data = tcr, x = axis1, y =axis3 , hue = list(map(str.lower,md_df_st_[col])), ax = ax3)


            else:
                fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3, color = list(map(str.lower,md_df_st_[col])) )
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , hue = list(map(str.lower,md_df_st_[col])), ax = ax1)
                sns.scatterplot(data = tcr, x = axis2, y =axis3 , hue = list(map(str.lower,md_df_st_[col])), ax = ax2)
                sns.scatterplot(data = tcr, x = axis1, y =axis3 , hue = list(map(str.lower,md_df_st_[col])), ax = ax3)

        # color with scores and metadata
        elif len(list(labels)) > 0  and not meta_data.empty:
            if clus_method in cluster_methods[1:]:
                filter = ['None', clus_method]
                filter.extend(md_df_st_.columns)
            else:
                filter = md_df_st_.columns.insert(0,'None')

            col = st.selectbox('Color by:', options= filter)
            if col == "None":
                fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3)
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , ax = ax1)
            elif col == clus_method:
                fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3, color = labels)
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , ax = ax1)
            else:
                fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3, color = list(map(str.lower,md_df_st_[col])))
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , hue = list(map(str.lower,md_df_st_[col])), ax = ax1)
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , hue = list(map(str.lower,md_df_st_[col])), ax = ax2)
                sns.scatterplot(data = tcr, x = axis1, y =axis2 , hue = list(map(str.lower,md_df_st_[col])), ax = ax3)

        else:
            fig = px.scatter_3d(tcr, x=axis1, y=axis2, z = axis3)
            sns.scatterplot(data = tcr, x = axis1, y =axis2 , ax = ax1)
        fig.update_traces(marker=dict(size=4))

        if selected_samples_idx:
            tt = tcr.iloc[selected_samples_idx,:]
            fig.add_scatter3d(x = tt.loc[:,axis1], y = tt.loc[:,axis2],
                              z = tt.loc[:,axis3], mode ='markers', marker = dict(size = 5, color = 'black'),
                              name = 'selected samples')
        
        plt.savefig("./Report/Figures/test.png")
        st.plotly_chart(fig, use_container_width=True)

        if labels:
            num_clusters = len(np.unique(labels))

            custom_color_palette = px.colors.qualitative.Plotly[:num_clusters]
            color_discrete_sequence=custom_color_palette

            # Créer et exporter le graphique Axe1-Axe2 en PNG
            fig_axe1_axe2 = px.scatter(tcr, x=axis1, y=axis2, color=labels if list(labels) else None, color_discrete_sequence=custom_color_palette)
            fig_axe1_axe2.update_layout(title='Axe1-Axe2')
            fig_axe1_axe2.update_traces(marker=dict(size=4))
            fig_axe1_axe2.write_image("./Report/Figures/plot_axe1_axe2.png")


            # Créer et exporter le graphique Axe1-Axe3 en PNG
            fig_axe1_axe3 = px.scatter(tcr, x=axis1, y=axis3, color=labels if list(labels) else None, color_discrete_sequence=custom_color_palette)
            fig_axe1_axe3.update_layout(title='Axe1-Axe3')
            fig_axe1_axe3.update_traces(marker=dict(size=4))
            fig_axe1_axe3.write_image("./Report/Figures/plot_axe1_axe3.png")


            # Créer et exporter le graphique Axe2-Axe3 en PNG
            fig_axe2_axe3 = px.scatter(tcr, x=axis2, y=axis3, color=labels if list(labels) else None, color_discrete_sequence=custom_color_palette)
            fig_axe2_axe3.update_layout(title='Axe2-Axe3')
            fig_axe2_axe3.update_traces(marker=dict(size=4))
            fig_axe2_axe3.write_image("./Report/Figures/plot_axe2_axe3.png")


if not spectra.empty:
    if dim_red_method == dim_red_methods[1] or dim_red_method == dim_red_methods[3]:
        with loadings:
            st.write('Loadings plot')
            p = dr_model.loadings_
            freq = pd.DataFrame(colnames, index=p.index)
            
            
            
            if test =='.dx':
                if meta_data.loc[:,'xunits'][0] == '1/cm':
                    freq.columns = ['Wavenumber (1/cm)']
                    xlab = "Wavenumber (1/cm)"
                    inv = 'reversed'
                else:
                    freq.columns = ['Wavelength (nm)']
                    xlab = 'Wavelength (nm)'
                    inv = None
                    
            else:
                freq.columns = ['Wavelength/Wavenumber']
                xlab = 'Wavelength/Wavenumber'
                inv = None
                
            pp = pd.concat([p, freq], axis=1)
            #########################################
            df1 = pp.melt(id_vars=freq.columns)
            fig = px.line(df1, x=freq.columns, y='value', color='variable', color_discrete_sequence=px.colors.qualitative.Plotly)
            fig.update_layout(legend=dict(x=1, y=0, font=dict(family="Courier", size=12, color="black"),
                                        bordercolor="black", borderwidth=2))
            fig.update_layout(xaxis_title = xlab,yaxis_title = "Intensity" ,xaxis = dict(autorange= inv))

            
            st.plotly_chart(fig, use_container_width=True)
            

            # Export du graphique
            img = pio.to_image(fig, format="png")
            with open("./Report/figures/graphe_loadings.png", "wb") as f:
                f.write(img)
#############################################################################################################
    if dim_red_method == dim_red_methods[1]:
        with influence:
            st.write('Influence plot')

            # Laverage
            Hat =  t.to_numpy() @ np.linalg.inv(np.transpose(t.to_numpy()) @ t.to_numpy()) @ np.transpose(t.to_numpy())
            leverage = np.diag(Hat) / np.trace(Hat)
            tresh3 = 2 * t.shape[1]/t.shape[0]
            # Loadings
            p = pd.concat([dr_model.loadings_.loc[:,axis1], dr_model.loadings_.loc[:,axis2], dr_model.loadings_.loc[:,axis3]], axis = 1)
            # Matrix reconstruction
            xp = np.dot(t,p.T)
            # Q residuals: Q residuals represent the magnitude of the variation remaining in each sample after projection through the model
            residuals = np.diag(np.subtract(xc.to_numpy(), xp)@ np.subtract(xc.to_numpy(), xp).T)
            tresh4 = sc.stats.chi2.ppf(0.05, df = 3)

            # color with metadata
            if not meta_data.empty and clus_method:
                if col == "None":
                    l1 = ["Samples"]* t.shape[0]

                elif col == clus_method:
                    l1 = labels
                
                else:
                    l1 = list(map(str.lower,md_df_st_[col]))

            elif meta_data.empty and clus_method:                        
                l1 = labels

            elif meta_data.empty and not clus_method:
                l1 = ["Samples"]* t.shape[0]
            
            elif not meta_data.empty and not clus_method:
                l1 = list(map(str.lower,md_df_st_[col]))

                    
                    
            fig = px.scatter(x = leverage, y = residuals, color = l1)
            fig.add_vline(x = tresh3, line_width = 1, line_dash = 'solid', line_color = 'red')
            fig.add_hline(y=tresh4, line_width=1, line_dash='solid', line_color='red')
            fig.update_layout(xaxis_title="Leverage", yaxis_title = "Residuals")

            out3 = leverage > tresh3
            out4 = residuals > tresh4

            for i in range(t.shape[0]):
                if out3[i]:
                    if not meta_data.empty:
                        ann =  meta_data.loc[:,'name'][i]
                    else:
                        ann = t.index[i]
                    fig.add_annotation(dict(x = leverage[i], y = residuals[i], showarrow=True, text = ann,
                                xanchor = 'auto', yanchor = 'auto'))
                
            st.plotly_chart(fig, use_container_width = True)
            img = pio.to_image(fig, format="png")
            with open("./Report/figures/graphe_influence.png", "wb") as f:
                f.write(img)


        with hotelling:
            st.write('T²-Hotelling vs Q residuals plot')
            # Hotelling
            hotelling  = t.var(axis = 1)
            # Q residuals: Q residuals represent the magnitude of the variation remaining in each sample after projection through the model
            residuals = np.diag(np.subtract(xc.to_numpy(), xp)@ np.subtract(xc.to_numpy(), xp).T)

            I = t.shape[0]
            fcri = sc.stats.f.isf(0.05, 3, I)
            tresh0 = (3 * (I ** 2 - 1) * fcri) / (I * (I - 3))
            tresh1 = sc.stats.chi2.ppf(0.05, df = 3)
            
            fig = px.scatter(t, x = hotelling, y = residuals, color = l1)
            fig.update_layout(xaxis_title="T²",yaxis_title="Q-Residuals")
            fig.add_vline(x=tresh0, line_width=1, line_dash='solid', line_color='red')
            fig.add_hline(y=tresh1, line_width=1, line_dash='solid', line_color='red')

            out0 = hotelling > tresh0
            out1 = residuals > tresh1

            
            for i in range(t.shape[0]):
                if out0[i]:
                    if not meta_data.empty:
                        ann =  meta_data.loc[:,'name'][i]
                    else:
                        ann = t.index[i]
                    fig.add_annotation(dict(x = hotelling[i], y = residuals[i], showarrow=True, text = ann,
                                xanchor = 'auto', yanchor = 'auto'))
                
            st.plotly_chart(fig, use_container_width=True)
            fig.write_image("./Report/figures/graphe_hotelling.png", format="png")
            #st.write()
            #st.write()
Nb_ech = str(tcr.shape[0])
nb_clu = str(sam1.shape[0])
Ac_Km = ['Spectra_Plot.png', 'Elbow.png', 'graphe_loadings.png', 'plot_axe1_axe2.png', 'plot_axe1_axe3.png', 'plot_axe2_axe3.png', 'graphe_hotelling.png', 'graphe_influence.png']

# Streamlit container
with st.container():
    if st.button("Download report"):
        if test == '.csv':
            if dim_red_method == dim_red_methods[1] and clus_method == cluster_methods[1]:
                latex_report = report.report(sam, tcr, Nb_ech, nb_clu, 'sample', Ac_Km, 'csv', 'kmeans')
                report.compile_latex()
            elif dim_red_method == dim_red_methods[1] and clus_method == cluster_methods[2]:
                latex_report = report.report(sam, tcr, Nb_ech, nb_clu, 'sample', Ac_Km, 'csv', 'hdb')
                report.compile_latex()
            elif dim_red_method == dim_red_methods[1] and clus_method == cluster_methods[3]:
                latex_report = report.report(sam, tcr, Nb_ech, nb_clu, 'sample', Ac_Km, 'csv', 'AP')
                report.compile_latex()
        else:
            latex_report = report.report(sam, 'sample', 'dx')
            report.compile_latex()
    else:
        pass