jdti 0.1.7

No project description provided

JDtI – Python library for scRNAseq/RNAseq data analysis

Author: Jakub Kubiś

Institute of Bioorganic Chemistry
Polish Academy of Sciences
Laboratory of Single Cell Analyses

Description

JDtI (JDataIntegration) is a Python library for data integration and advanced post-processing of single-cell datasets.

JDtI enables basic quality control steps such as control of cells per cluster, number of genes per cell, and more advanced tasks like subclustering, integration, and wide visualization. In this approach, we do not drop the cell information during separate set analyses; instead, we use previous cluster cell lineage information for integrating data based on cluster markers and data harmonization. After integration, it is possible to visualize cell interactions and correlations in many ways, including cell distance, correlations, and more.

Despite this, it is also able to conduct DEG analysis between sets, selected cells, or grouped cells, and visualize the results on UMAP, volcano plots, and regression plots comparing pairs of cells. It is very powerful for more advanced analyses focusing on specific issues within the data that may not be discovered in basic analyses.

Additionally, JDtI offers many functions for data visualization and processing within clean visual outputs, such as volcano plots, gene expression analysis of different data types, clustering, heatmaps, and more.

It is compatible with various sequencing approaches, including scRNA-seq and bulk RNA-seq, and supports interoperability with tools such as Seurat, Scanpy, and other bioinformatics frameworks using the 10x sparse matrix format as input. More details about the available functions can be found in the Documentation and Example Usage section on GitHub.

📚 Table of Contents

• Installation
• Documentation
• Example usage
  • 1. Basic functions
    • 1.1. Loading functions
    • 1.2. Loading data
    • 1.3. Features finding
    • 1.4. Names finding
    • 1.5. Data reducing
    • 1.6. Data averaging and occurrence counting
    • 1.7. Difference counting (DEG) and visualization
    • 1.8. Features visualization
    • 1.9. Relation visualization
  • 2. Data clustering
    • 2.1. Loading class and helper function
    • 2.2. Loading data
    • 2.3. Initialize Clustering class
    • 2.4. Performing PCA
    • 2.5. Knee plot of PC
    • 2.6. Harmonize data (Harmony)
    • 2.7. Find clusters on PC
    • 2.8. Perform UMAP
    • 2.9. Knee plot of UMAP
    • 2.10. Find clusters on UMAP
    • 2.11. Visualization of names on UMAP reduced space
    • 2.12. Visualization of feature level on UMAP reduced space
  • 3. Data integration
    • 3.1. Loading class and helper functions
    • 3.2. Initialize COMPsc class
    • 3.3. Loading data
    • 3.4. Normalize data
    • 3.5. Gene amount thresholds visualisation
    • 3.6. Gene amount thresholds adjustment
    • 3.7. Sample reduction
    • 3.8. Samples / cells amount visualisation
    • 3.9. Sample / cell amount threshold
    • 3.10. Calculation of differential markers for integration
    • 3.11. Calculation of samples / cells similarity factors
    • 3.12. Similarity visualisation (correlation × distance)
    • 3.13. Similarity visualisation (spatial distance)
    • 3.14. Similarity visualisation (sample to sample)
    • 3.15. Clustering features estimation
    • 3.16. Performing PCA
    • 3.17. Knee plot of PC
    • 3.18. Harmonize data (Harmony)
    • 3.19. Perform UMAP
    • 3.20. Visualization of names / sets on UMAP reduced space
    • 3.21. Visualization of feature level on UMAP reduced space
    • 3.22. De novo clustering - Performing PCA
    • 3.23. De novo clustering - Knee plot of PC
    • 3.24. De novo clustering - Harmonize data (Harmony)
    • 3.25. De novo clustering - Find clusters on PC
    • 3.26. De novo clustering - Perform UMAP
    • 3.27. De novo clustering - Knee plot of UMAP
    • 3.28. De novo clustering - Find clusters on UMAP
    • 3.29. De novo clustering - Visualization of clusters on UMAP reduced space
    • 3.30. Differential expression analysis
    • 3.31. Volcano plot
    • 3.32. Scatter plot
    • 3.33. Visualization of feature level on UMAP reduced space
    • 3.34. Sample / cell composition calculation
    • 3.35. Composition - pie plot
    • 3.36. Composition - bar plot
    • 3.37. Getting data
    • 3.38. Saving sparse data
    • 3.39. Saving project
    • 3.40. Loading project
  • 4. Data subclustering
    • 4.1. Loading class
    • 4.2. Initialize COMPsc class
    • 4.3. Loading data
    • 4.4. Select cluster and features for subclustering
    • 4.5. Prepare subclustering
    • 4.6. Define subclusters
    • 4.7. Visualize subclusters features
    • 4.8. Adjust subclusters and check features
    • 4.9. Calculate DEG for subclusters and visualize
    • 4.10. Confirm subclusters and associate with data

Installation

pip install jdti

Documentation

Documentation for classes and functions is available here 👉 Documentation 📄

Example usage

1. Basic functions

1.1. Loading functions

from jdti import *

1.2. Loading data

# load sparse matrix as pd.DataFrame data with creating metadata
data, metadata = load_sparse(path = 'data/set1', name = 'set1')

#load data frame from different data type (.tsv, .txt, .tsv)
data = pd.read_csv('example_data.csv')

# load data from .h5 or other data types and transform to pandas data frame

• Data [features (eg. genes) x sample (eg. cells)]
• Metadata [columns['cell_names', 'sets']]:
  • cell_names – sample names corresponding to the columns of Data
  • sets – the assignment of each sample to a given dataset, aligned with Data

1.3. Features finding

features = find_features(data, features =['KIT', 'MC1', 'EDNRB', 'PAX3'])

• Not found the MC1 feature name, so the potential names are provided

features = find_features(data, features =['KIT', 'MC1R', 'EDNRB', 'PAX3'])

• All feature names have been found

1.4. Names finding

names = find_names(data, names = ['0', '1', '2','10', '1&'])

• As same as in case of 'Features finding'

1.5. Data reducing

# data reducing on found features and names

data_reduced = reduce_data(data,
                features = features['included'],
                names = names['included'])

• return data with selected features & names

1.6. Data averaging and occurrence counting

avg_reduced = average(data_reduced)
occ_reduced = occurrence(data_reduced)

• returns the average or occurrence values computed across all columns that share the same name

1.7. Difference counting (DEG) and visualization

# creating group dict for compare samples
compare_dict = {'g1':['0', '1'],
                'g2':['2','10']}


deg_df = calc_DEG(data, 
             metadata_list  = None, 
             entities = compare_dict, 
             sets = None, 
             min_exp = 0, 
             min_pct = 0.1, 
             n_proc =10)

# DEG visualization with volcano plot
fig = volcano_plot(deg_df, 
                 p_adj = True, 
                 top = 25, 
                 p_val = 0.05, 
                 lfc = 0.25, 
                 standard_scale = False, 
                 rescale_adj = True, 
                 image_width = 12, 
                 image_high = 12)

fig.savefig('volcano.jpeg', dpi=300, bbox_inches='tight')

• DEG data:
  • feature – Name of the studied feature
  • p_val – P-value (Mann–Whitney) for the studied feature comparing the valid_group to all other groups in the analysis
  • pct_valid – Percentage of positive (>0) values for the studied feature in the valid_group*
  • pct_ctrl – Percentage of positive (>0) values for the studied feature in all other groups
  • avg_valid – Average value of the studied feature in the valid_group
  • avg_ctrl – Average value of the studied feature in the remaining groups
  • sd_valid – Standard deviation of the studied feature in the valid_group*
  • sd_ctrl – Standard deviation of the studied feature in the remaining groups
  • esm – Cohen’s d effect size metric
  • valid_group – Name of the sample or group belonging to the valid_group
  • adj_pval – Benjamini–Hochberg adjusted p-value
  • FC – Fold change between the averaged valid_group samples and the averaged remaining samples
  • log(FC) – Log₂-transformed fold change
  • norm_diff – Direct difference between the averaged valid_group value and the averaged value of the remaining groups

• Volcano plot – Visualization of differentially expressed genes (DEGs) between two groups

1.8. Features visualization

top_10 = deg_df.sort_values(
    ['p_val', 'esm', 'log(FC)'], 
    ascending=[True, False, False]).head(10)

data_scatter = reduce_data(data,
                features = list(set(top_10['feature'])),
                names = names['included'])



avg = average(data_scatter)
occ = occurrence(data_scatter)


fig = features_scatter(expression_data = avg, 
                     occurence_data = occ,
                     features = None, 
                     metadata_list = None, 
                     colors = 'viridis', 
                     hclust = 'complete', 
                     img_width = 8, 
                     img_high = 5, 
                     label_size = 10, 
                     size_scale = 100,
                     y_lab = 'Genes', 
                     legend_lab = 'log(CPM + 1)',
                     bbox_to_anchor_scale = 25,
                     bbox_to_anchor_perc=(0.91, 0.55),
                     bbox_to_anchor_group=(1.01, 0.4))

fig.savefig('scatter.jpeg', dpi=300, bbox_inches='tight')

• Scatter plot – Displays expression relationships of DEGs across groups or individual samples

1.9. Relation visualization

fig = development_clust(data = avg, 
                      method = 'ward',
                      img_width = 5,
                      img_high = 5)

fig.savefig('development.jpeg', dpi=300, bbox_inches='tight')

• Development plot – A dendrogram showing sample similarity based on the expression features generated using hierarchical clustering

2. Data clustering

2.1. Loading class and helper functions

from jdti import Clustering, load_sparse

2.2. Loading data

# load sparse matrix as pd.DataFrame data with creating metadata
data, metadata = load_sparse(path = 'set1', name = 'set1')

#load data frame from different data type (.tsv, .txt, .tsv)
data = pd.read_csv('example_data.csv')

# load data from .h5 or other data types and transform to pandas data frame

• Data [features (eg. genes) x sample (eg. cells)]
• Metadata [columns['cell_names', 'sets']]:
  • cell_names – sample names corresponding to the columns of Data
  • sets – the assignment of each sample to a given dataset, aligned with Data

2.3. Initialize Clustering class

clusters = Clustering.add_data_frame(data, metadata)

# attributes with inputed data and metadata
clusters.clustering_data
clusters.clustering_metadata

2.4. Performing PCA

fig1 = clusters.perform_PCA(pc_num=50, width=8, height=6)

fig1.savefig('clus_PCA.jpeg', dpi=300, bbox_inches='tight')

2.5. Knee plot of PC

fig2 = clusters.knee_plot_PCA(width=8, height=6)

fig2.savefig('clus_PCA_knee.jpeg', dpi=300, bbox_inches='tight')

2.6. Harmonize data (Harmony)

# if more than one dataset (provided in metadata) harmonization process can be used
clusters.harmonize_sets()

2.7. Find clusters on PC

fig3 = clusters.find_clusters_PCA(pc_num=50, eps=0.5, min_samples=10, width=8, height=6, harmonized=False)

clusters.return_clusters(clusters='pca')

fig3.savefig('clus_PCA_clusters.jpeg', dpi=300, bbox_inches='tight')

• No cluster detected in linear reduced space

2.8. Perform UMAP

clusters.perform_UMAP(factorize=True, umap_num=50, pc_num=5, harmonized=False)

2.9. Knee plot of UMAP

fig4 = clusters.knee_plot_umap(eps=0.5, min_samples=10)

fig4.savefig('clus_UMAP_knee.jpeg', dpi=300, bbox_inches='tight')

2.10. Find clusters on UMAP

fig5 = clusters.find_clusters_UMAP(
            umap_n=2, 
            eps=0.5, 
            min_samples=10, 
            width=8, 
            height=6)

clusters.return_clusters(clusters='umap')

fig5.savefig('clus_UMAP_clusters.jpeg', dpi=300, bbox_inches='tight')

2.11. Visualization of names on UMAP reduced space

fig6 = clusters.UMAP_vis(names_slot='cell_names', set_sep=True, point_size=0.6)

fig6.savefig('clus_UMAP_names_vis.jpeg', dpi=300, bbox_inches='tight')

fig7 = clusters.UMAP_vis(names_slot='UMAP_clusters', set_sep=True, point_size=0.6)

fig7.savefig('clus_UMAP_clusters_vis.jpeg', dpi=300, bbox_inches='tight')

2.12. Visualization of feature level on UMAP reduced space

fig8 = clusters.UMAP_feature(
            feature_name = '001G000700', 
            features_data=None, 
            point_size=0.6)

fig8.savefig('clus_UMAP_features.jpeg', dpi=300, bbox_inches='tight')

3. Data integration

3.1. Loading class and helper functions

from jdti import COMPsc, volcano_plot

3.2. Initialize COMPsc class

import os

jseq_object = COMPsc.project_dir(os.getcwd(), ['set1', 'set2'])

3.3. Loading data

jseq_object.load_sparse_from_projects(normalized_data=True)

# attributes with inputed data and metadata
jseq_object.input_metadata

# if normalized_data=False
jseq_object.input_data

# if normalized_data=True
jseq_object.normalized_data

3.4. Normalize data

# use if inputed_data is count data

jseq_object.normalize_counts(normalize_factor = 100000, 
                            log_transform = True)

3.5. Gene amount thresholds visualisation

fig = jseq_object.gene_histograme(bins=100)

fig.savefig('int_hist_genes.jpeg', dpi=300, bbox_inches='tight')

3.6. Gene amount thresholds adjustment

jseq_object.gene_threshold(min_n = 500, max_n = 6000)

fig = jseq_object.gene_histograme(bins=100)

fig.savefig('int_hist_genes_reduced.jpeg', dpi=300, bbox_inches='tight')

3.7. Sample reduction

jseq_object.reduce_cols(reg = 'xylem / phloem', inc_set = False)

fig = jseq_object.gene_histograme(bins=100)

fig.savefig('int_hist_genes_reduced_names.jpeg', dpi=300, bbox_inches='tight')

3.8. Samples / cells amount visualisation

fig = jseq_object.cell_histograme(name_slot = 'cell_names')

fig.savefig('int_hist_cell_names.jpeg', dpi=300, bbox_inches='tight')

3.9. Sample / cell amount threshold

jseq_object.cluster_threshold(min_n = 50, name_slot = 'cell_names')

fig = jseq_object.cell_histograme(name_slot = 'cell_names')

fig.savefig('int_hist_cell_names_reduced.jpeg', dpi=300, bbox_inches='tight')

3.10. Calculation of differential markers for integration

jseq_object.calculate_difference_markers(min_exp = 0, 
                                         min_pct = 0.25, 
                                         n_proc=10, 
                                         force = False)

3.11. Calculation of samples / cells similarity factors

jseq_object.estimating_similarity(method = 'pearson', 
                                  p_val = 0.05,
                                  top_n = 25)

3.12. Similarity visualisation (correlation x distance)

fig1 = jseq_object.similarity_plot(split_sets = True, 
                                 set_info = True,
                                 cmap='seismic', 
                                 width = 8, height = 5)
    
fig1.savefig('int_sim_plot_top.jpeg', dpi=300, bbox_inches='tight')

3.13. Similarity visualisation (spatial distance)

fig2 = jseq_object.spatial_similarity(set_info= True, 
                                     bandwidth = 1,
                                        n_neighbors = 6,  
                                        min_dist = 0.5, 
                                        legend_split = 2, 
                                        point_size = 200, 
                                        spread=1.0,              
                                        set_op_mix_ratio=1.0,    
                                        local_connectivity=1,    
                                        repulsion_strength=1.0,  
                                        negative_sample_rate=5,  
                                        width = 6, 
                                        height = 4)



fig2.savefig('int_sim_plot_map_top.jpeg', dpi=300, bbox_inches='tight')

3.14. Similarity visualisation (sample to sample)

fig3 = jseq_object.cell_regression( 
                    cell_x = 'phloem_4', 
                    cell_y = 'phloem_3',
                    set_x = 'set1', 
                    set_y = 'set2',
                    threshold = 2, 
                    image_width = 12, 
                    image_high = 7, 
                    color = 'green')

fig3.savefig('int_sim_reg.jpeg', dpi=300, bbox_inches='tight')

3.15. Clustering features estimation

jseq_object.clustering_features(name_slot = 'cell_names', 
                                features_list = None, 
                                p_val = 0.05,
                                top_n = 25,
                                adj_mean = True,
                                beta = 0.2)

3.16. Performing PCA

fig4 = jseq_object.perform_PCA(pc_num = 50)

fig4.savefig('int_pca.jpeg', dpi=300, bbox_inches='tight')

3.17. Knee plot of PC

fig5 = jseq_object.knee_plot_PCA()

fig5.savefig('int_pca_knee.jpeg', dpi=300, bbox_inches='tight')

3.18. Harmonize data (Harmony)

# if more than one dataset (provided in metadata) harmonization 
# process for integration element can be used

jseq_object.harmonize_sets()

3.19. Perform UMAP

jseq_object.perform_UMAP(factorize=False, 
                        umap_num = 2, 
                        pc_num = 15, 
                        harmonized = True)

3.20. Visualization of names / sets on UMAP reduced space

fig6 = jseq_object.UMAP_vis( 
             names_slot = 'cell_names', 
             set_sep = True,
             point_size = 1,
             font_size = 10,
             legend_split_col = 1,
             width = 8,
             height = 6,
             inc_num = True)


fig6.savefig('int_umap.jpeg', dpi=300, bbox_inches='tight')

fig7 = jseq_object.UMAP_vis( 
             names_slot = 'sets', 
             set_sep = True,
             point_size = 1,
             font_size = 6,
             legend_split_col = 1,
             width = 8,
             height = 6,
             inc_num = False)

fig7.savefig('int_umap_sets.jpeg', dpi=300, bbox_inches='tight')

3.21. Visualization of feature level on UMAP reduced space

fig8 = jseq_object.UMAP_feature( 
             features_data = jseq_object.get_data(set_info = False) ,
             feature_name = '001G069799',
             point_size = 0.8,
             font_size = 6,
             width = 8,
             height = 6,
             palette = 'light')

fig8.savefig('int_umap_feature.jpeg', dpi=300, bbox_inches='tight')

3.22. De novo clustering - Performing PCA

fig9 = jseq_object.perform_PCA(pc_num = 50)

fig9.savefig('int_pca_clusters.jpeg', dpi=300, bbox_inches='tight')

3.23. De novo clustering - Knee plot of PC

fig10 = jseq_object.knee_plot_PCA()

fig10.savefig('int_pca_knee_clusters.jpeg', dpi=300, bbox_inches='tight')

3.24. De novo clustering - Harmonize data (Harmony)

# if more than one dataset (provided in metadata) harmonization 
# process for integration element can be used

jseq_object.harmonize_sets()

3.25. De novo clustering - Find clusters on PC

fig11 = jseq_object.find_clusters_PCA(
                        pc_num = 10, 
                        eps = 3, 
                        min_samples = 20, 
                        harmonized = True)

fig11.savefig('int_pca_clusters_find.jpeg', dpi=300, bbox_inches='tight')

3.26. De novo clustering - Perform UMAP

jseq_object.perform_UMAP(
                factorize=False, 
                umap_num = 10, 
                pc_num = 10, 
                harmonized = True)

3.27. De novo clustering - Knee plot of UMAP

fig12 = jseq_object.knee_plot_umap(eps = 0.5, min_samples = 10)

fig12.savefig('int_umap_knee.jpeg', dpi=300, bbox_inches='tight')

3.28. De novo clustering - Find clusters on UMAP

fig13 = jseq_object.find_clusters_UMAP(umap_n = 4,
                              eps = 0.5,
                              min_samples = 10,
                              width = 8,
                              height = 6)

fig13.savefig('int_umap_clusters.jpeg', dpi=300, bbox_inches='tight')

3.29. De novo clustering - Visualization of clusters on UMAP reduced space

fig14 = jseq_object.UMAP_vis( 
             names_slot = 'UMAP_clusters', 
             set_sep = True,
             point_size = 1,
             font_size = 10,
             legend_split_col = 1,
             width = 8,
             height = 6,
             inc_num = True)

fig14.savefig('int_umap_clusters_get_clusters_sep_set.jpeg', dpi=300, bbox_inches='tight')

fig15 = jseq_object.UMAP_vis( 
             names_slot = 'UMAP_clusters', 
             set_sep = False,
             point_size = 1,
             font_size = 10,
             legend_split_col = 1,
             width = 8,
             height = 6,
             inc_num = True)


fig15.savefig('int_umap_clusters_get_clusters_sep_set_not.jpeg', dpi=300, bbox_inches='tight')

3.30. Differential expression analysis

# if you want calculate markers for de novo clusters change metadata 'cell_names'
# jseq_object.input_metadata['cell_names'] = jseq_object.input_metadata['UMAP_clusters']

# calculation stats for all cells in 'cell_anmes'
stats = jseq_object.statistic(cells='All', sets=None, min_exp=0, min_pct=0.25, n_proc=10)

# calculation stats for all clusters in 'UMAP_clusters'
cell_names = jseq_object.input_metadata['cell_names'] # save cell_names for the future
jseq_object.input_metadata['cell_names'] = jseq_object.input_metadata['UMAP_clusters']
stats = jseq_object.statistic(cells='All', sets=None, min_exp=0, min_pct=0.25, n_proc=10)

# calculation stats for all sets in 'sets',

stats = jseq_object.statistic(cells=None, sets='All', min_exp=0, min_pct=0.25, n_proc=10)

• DEG data:
  • feature – Name of the studied feature
  • p_val – P-value (Mann–Whitney) for the studied feature comparing the valid_group to all other groups in the analysis
  • pct_valid – Percentage of positive (>0) values for the studied feature in the valid_group*
  • pct_ctrl – Percentage of positive (>0) values for the studied feature in all other groups
  • avg_valid – Average value of the studied feature in the valid_group
  • avg_ctrl – Average value of the studied feature in the remaining groups
  • sd_valid – Standard deviation of the studied feature in the valid_group*
  • sd_ctrl – Standard deviation of the studied feature in the remaining groups
  • esm – Cohen’s d effect size metric
  • valid_group – Name of the sample or group belonging to the valid_group
  • adj_pval – Benjamini–Hochberg adjusted p-value
  • FC – Fold change between the averaged valid_group samples and the averaged remaining samples
  • log(FC) – Log₂-transformed fold change
  • norm_diff – Direct difference between the averaged valid_group value and the averaged value of the remaining groups

3.31. Volcano plot

# select 'valid_group' for volcano plot create; this case 'set1' or 'set2' 
stats_volcano = stats[stats['valid_group'] == 'set1']

fig16 = volcano_plot(deg_data = stats_volcano,
                  p_adj = True,
                  top = 10,
                  p_val = 0.05,
                  lfc = 0.25,
                  standard_scale = False,
                  rescale_adj = True,
                  image_width = 12,
                  image_high = 13)

fig16.savefig('int_volcano.jpeg', dpi=300, bbox_inches='tight')

• Volcano plot – Visualization of differentially expressed genes (DEGs) between two groups

3.32. Scatter plot

stats_5 = stats.sort_values(['valid_group', 'esm', 'log(FC)'], ascending=[True, False, False]).groupby('valid_group').head(5)

fig17 = jseq_object.scatter_plot(
                 names = None,
                 features = list(set(stats_5['feature'])),
                 name_slot = 'cell_names',
                 scale = True,
                 colors = 'viridis', 
                 hclust = 'complete', 
                 img_width  = 15, 
                 img_high  = 3, 
                 label_size = 10, 
                 size_scale = 200,
                 y_lab = 'Genes', 
                 legend_lab = 'log(CPM + 1)',
                 set_box_size = 5,
                 set_box_high = 0.1,
                 bbox_to_anchor_scale = 25,
                 bbox_to_anchor_perc=(0.90, 0.5),
                 bbox_to_anchor_group=(0.9, 0.3))


fig17.savefig('int_scatter_DEG.jpeg', dpi=300, bbox_inches='tight')

3.33. Visualization of feature level on UMAP reduced space

fig18_1 = jseq_object.UMAP_feature(
            feature_name = 'T011200', 
            features_data=jseq_object.normalized_data, 
            point_size=0.6)

fig18_1.savefig('int_umap_clusters_feature_T011200.jpeg', dpi=300, bbox_inches='tight')

fig18_2 = jseq_object.UMAP_feature(
            feature_name = '001G158400', 
            features_data=jseq_object.normalized_data, 
            point_size=0.6)

fig18_2.savefig('int_umap_clusters_feature_001G158400.jpeg', dpi=300, bbox_inches='tight')

3.34. Sample / cell composition calculation

import re

jseq_object.data_composition( 
                     features_count = list(set([re.sub(r'_.*$', '',x) for x in list(set(jseq_object.input_metadata['cell_names']))])), # get names without numbers for composition calculation
                     name_slot = 'cell_names',
                     set_sep = True
                     )

3.35. Composition - pie plot

fig19 = jseq_object.composition_pie( 
                    width = 6, 
                    height = 6, 
                    font_size = 15,
                    cmap  = "tab20",
                    legend_split_col = 1,
                    offset_labels = 0.5,
                    legend_bbox = (1.15, 0.95))


fig19.savefig('int_composition_pie.jpeg', dpi=300, bbox_inches='tight')

3.36. Composition - bar plot

fig20 = jseq_object.bar_composition( 
                    cmap = 'tab20b', 
                    width = 2, 
                    height = 6, 
                    font_size = 15,
                    legend_split_col = 1,
                    legend_bbox = (1.3, 1))


fig20.savefig('int_composition_bar.jpeg', dpi=300, bbox_inches='tight')

3.37. Getting data

met = jseq_object.input_metadata

# full data
data = jseq_object.get_data(set_info=True) 

# metadata
metadata = jseq_object.get_metadata()

# partial data
dt = jseq_object.get_partial_data(
                names=['phloem_6'], 
                features=['001G000700', '019G047850', '019G089366'], 
                name_slot='cell_names')

# more in documentation

3.38. Saving sparse data

# save data from slots in sparse matrix format

jseq_object.save_sparse(
                path_to_save = "data",
                name_slot: str = "cell_names",
                data_slot: str = "normalized",
)

3.39. Saving project

# save whole project with analyses and results

jseq_object.save_project(name = 'tree')

3.40. Loading project

# load project with all attributes & methodes

reoladed_project = COMPsc.load_project('tree.jpkl')

4. Data subclustering

4.1. Loading class

from jdti import COMPsc

4.2. Initialize COMPsc class

import os

jseq_object = COMPsc.project_dir(os.getcwd(), ['set2'])

4.3. Loading data

jseq_object.load_sparse_from_projects(normalized_data=True)

4.4. Select cluster and features for subclustering

jseq_object.calculate_difference_markers()

set(jseq_object.normalized_data.columns)

set_markers = jseq_object.var_data[jseq_object.var_data['valid_group'] == 'xylem / phloem_2 # set2']

set_markers = set_markers.sort_values(['esm', 'log(FC)'], ascending=[False, False]).head(5)

4.5. Prepare subclustering

jseq_object.subcluster_prepare(features = list(set_markers['feature']), 
                               cluster='xylem / phloem_2')

4.6. Define subclusters

fig1 = jseq_object.define_subclusters( 
                          umap_num = 5,
                          eps = 1.1, 
                          min_samples = 5,
                          n_neighbors = 5,  
                          min_dist = 0.1, 
                          spread = 1.0,              
                          set_op_mix_ratio = 1.0,    
                          local_connectivity = 1,    
                          repulsion_strength = 1.0,  
                          negative_sample_rate = 5,  
                          width = 8, 
                          height = 6)
  
fig1.savefig('sub_umap_clust.jpeg', dpi=300, bbox_inches='tight')

4.7. Visualize subclusters features

fig2 = jseq_object.subcluster_features_scatter(
                                        colors = 'viridis', 
                                        hclust = 'complete', 
                                        img_width = 3, 
                                        img_high = 5, 
                                        label_size = 6, 
                                        size_scale = 70,
                                        y_lab = 'Genes', 
                                        legend_lab = 'normalized')
    
fig2.savefig('sub_scatter_clust_genes.jpeg', dpi=300, bbox_inches='tight')

4.8. Adjust subclusters and check features

# if subclusters are very similar to each other 
# you can use merging subclusters and visualise again

mapping = {
    "old_name": ["1", "2"],
    "new_name": ["1", "1"]
}

jseq_object.rename_subclusters(mapping)

fig3 = jseq_object.subcluster_features_scatter(
                                        colors = 'viridis', 
                                        hclust = 'complete', 
                                        img_width = 3, 
                                        img_high = 5, 
                                        label_size = 6, 
                                        size_scale = 70,
                                        y_lab = 'Genes', 
                                        legend_lab = 'normalized')

fig3.savefig('sub_scatter_clust_genes_reduced.jpeg', dpi=300, bbox_inches='tight')

4.9. Calculate DEG for subclusters and visualize

fig4 = jseq_object.subcluster_DEG_scatter(
                                    top_n = 3,
                                    min_exp = 0, 
                                    min_pct = 0.1, 
                                    p_val = 0.05,
                                    colors = 'viridis', 
                                    hclust = 'complete', 
                                    img_width = 3, 
                                    img_high = 5, 
                                    label_size = 6, 
                                    size_scale = 70,
                                    y_lab = 'Genes', 
                                    legend_lab = 'normalized',
                                    n_proc=10)
    
fig4.savefig('sub_scatter_clust_genes_reduced_DEG.jpeg', dpi=300, bbox_inches='tight')

4.10. Confirm subclusters and associate with data

set(jseq_object.input_metadata['cell_names'])

Output before accept:
{'phloem_1', 'phloem_2', 'phloem_3', 'phloem_5', 'xylem / phloem_2', 'xylem / phloem_3', 'xylem_1', 'xylem_4', 'xylem_7', 'xylem_8'}

jseq_object.accept_subclusters()
    
set(jseq_object.input_metadata['cell_names'])

Output after accept: {'phloem_1', 'phloem_2', 'phloem_3', 'phloem_5', 'xylem / phloem_2.0', 'xylem / phloem_2.1', 'xylem / phloem_2.3', 'xylem / phloem_2.4', 'xylem / phloem_3', 'xylem_1', 'xylem_4', 'xylem_7', 'xylem_8'}

Have fun JBS