Single-cell RNA sequencing revealed organ-specific endothelial landscapes and intercellular networks in endocrine organs

## 1. Scope of This README This README documents only the materials currently available in the following paths: - `Figure1/` - `Figure2/` - `Figure3/` - `Figure4_5/` - `Figure6/` - `Figure7/` - `h5ad/` This file is intended to help editors and reviewers identify: - which processed `AnnData` objects are included, - which notebooks/scripts generated the representative analyses for each main figure, - which parts of the analysis are already documented inside this package. ## 2. Important Notes - Many notebooks still contain legacy absolute paths from the original workstation. For reviewer sharing, those paths should be updated to the local copies under `h5ad/`. ## 3. Preprocessing (Mouse Atlas) The mouse atlas preprocessing is documented in `Preprocess/` and proceeds as follows: ### 3-1. Read mapping — `Preprocess/00_cellranger_script.sh` - Tool: **Cell Ranger v7.1.0** - Reference transcriptome: `refdata-gex-mm10-2020-A` ### 3-2. Ambient RNA removal — `Preprocess/00_cellbender_script.sh` - Tool: **CellBender** `remove-background` (GPU mode, `--cuda`) - Input: `raw_feature_bc_matrix.h5` per sample ### 3-3. QC filtering and doublet detection — `Preprocess/02_scrublet.ipynb`, `Preprocess/03_Filtering_QC.ipynb` - Doublet scoring with **Scrublet** - Cells removed if: low UMI/gene counts, high gene counts (putative doublets), or high mitochondrial fraction ### 3-4. Batch correction and integration — `Preprocess/04_pre_batchcorr.ipynb`, `Preprocess/05_BBKNN.ipynb` - BGlobal batch correction: **BBKNN** across samples/organs - Endothelial subclustering and harmony integration: `Preprocess/06_sub_endo_all_harmony.ipynb` ## 4. Processed AnnData Inventory The main processed objects available inside `h5ad/` are: - `h5ad/SNU_cellbender_filtered_BBKNN_Anno.h5ad` - Integrated mouse dataset with broad cell type annotations. - `h5ad/03_SNU_cellbender_filtered_QC.h5ad` - QC-filtered mouse object used as a raw-count reference in downstream endothelial/capillary analyses. (raw counts) - `h5ad/FIg2_h5ad/subcluster_EC_harmony_Anno.h5ad` - Endothelial subset after integration/subclustering/annotation. - `h5ad/Fig3_h5ad/09_subcluster_capi_harmony_commo_speci.h5ad` - Capillary-only object with common/specialized capillary annotations. - `h5ad/Fig4_h5ad/for_pseudobulk_organotypic_onlycapi.h5ad` - Capillary-only object used for organotypic pseudobulk DEG and network analysis in Figure 4_5. - `h5ad/Fig4_h5ad/Fig4_capiilary_human.h5ad` - Human capillary object used for cross-species organ-specific marker comparisons in Figure 4. - `h5ad/Fig4_h5ad/Fig4_All_celltyp_human.h5ad` - Human all-cell-type object used in supplementary cross-species marker analyses. - `h5ad/Fig7_h5ad/FIg7_Thyroid_BBKNN_hvg3500_low_anno.h5ad` - Human thyroid(Normal+PTC) dataset after integration and low-level annotation. - `h5ad/Fig7_h5ad/FIg7_Thyroid_EC_subclustering_anno.h5ad` - Thyroid endothelial subclustering object. - `h5ad/Fig7_h5ad/for_cci/young.h5ad` - `h5ad/Fig7_h5ad/for_cci/old.h5ad` - `h5ad/Fig7_h5ad/for_cci/ptc.h5ad` - Condition-specific thyroid objects used for Cell Cell Interaction analyses in Figure 7. ## 5. Figure-by-Figure Analysis Map ### Figure 1. Global mouse atlas clustering and annotation Representative notebooks: - `Figure1/Fig1B_D.ipynb` - `Figure1/Fig1E.ipynb` - `Figure1/Supplementary_Fig1.ipynb` - `Figure1/Supplementary_Fig2.ipynb` - `Figure1/Supplementary_Fig3.ipynb` Representative analyses: - global clustering and annotation of the integrated mouse atlas, - UMAP visualization of organ and cell type structure, - assessment of marker gene expression based on UMAP visualization, - organ-by-cell type composition summaries. ### Figure 2. Endothelial annotation, regulon analysis, and pseudobulk comparison (in capillary 3-types) Representative notebooks: - `Figure2/01/Figure2_A_D.ipynb` - `Figure2/01/Figure2_E_F.ipynb` - `Figure2/01/Supplementary_Fig4.ipynb` - `Figure2/02_scenic/Figure2_G.ipynb` - `Figure2/02_scenic/Figure2_H.ipynb` - `Figure2/03_3type_pseudobulk/03_1_conti-fene vs ohter_ECtype.ipynb` - `Figure2/03_3type_pseudobulk/03_2_conti-non-fene vs ohter_ECtype.ipynb` - `Figure2/03_3type_pseudobulk/03_3_Sinusoida vs ohter_ECtype.ipynb` - `Figure2/03_3type_pseudobulk/03_4_GSEA_dotplot_Fig2J.ipynb` - `Figure2/03_3type_pseudobulk/03_5_3type_network_makeInput.ipynb` Main input object: - `h5ad/FIg2_h5ad/subcluster_EC_harmony_Anno.h5ad` Supporting raw-count reference: - `h5ad/03_SNU_cellbender_filtered_QC.h5ad` Representative analyses: - endothelial subcluster annotation, - identification of marker genes for the three capillary endothelial types, - SCENIC/regulon summary for the three capillary endothelial types, - pseudobulk differential comparisons across the three capillary endothelial types, - ORA (over-representation analysis using the hypergeometric test) was performed on DEGs for each capillary type, - Significant ORA terms were defined as nodes, and node pairs were filtered based on the Jaccard index of overlapping genes to generate input files for Cytoscape-style visualization. ### Figure 3. Capillary common/specialized states, regulon analysis, and NMF program analysis Representative notebooks/scripts: - `Figure3/01/00_1_Capillary_Figure3_A_H.ipynb` - `Figure3/01/Supplementary_Figure6_C.ipynb` - `Figure3/02_scenic/Figure3_I.ipynb` - `Figure3/03_NMF/Supplementary_Figure7_A_E.ipynb` - `Figure3/03_NMF/Supplementary_Figure7_F.ipynb` - `Figure3/03_NMF/run_nmf.py` Main input object: - `h5ad/Fig3_h5ad/09_subcluster_capi_harmony_commo_speci.h5ad` Representative analyses: - clustering-based annotation of common and organ-specialized capillary states within the capillary subset, - visualization of capillary state UMAPs, organ distributions, organ proportion evenness across clusters, and capillary state DEGs, - SCENIC/regulon analysis of capillary states, - comparison of clustering-based capillary state annotations with NMF-identified gene expression programs. ### Figure 4 and Figure 5. Organ-specific capillary DEGs, organotypic pseudobulk DEG, ORA reulsts network, and mouse-human comparison Representative notebooks/scripts: - `Figure4_5/01_DEG_jaccard_sim/Figure4_A_B.ipynb` - `Figure4_5/02_Pseudobulk_network/01_EC_organotypic_run_pseudobulk_DEG.py` - `Figure4_5/02_Pseudobulk_network/02_EC_conserved_marker.ipynb` - `Figure4_5/02_Pseudobulk_network/03_remove_nonEC_enriched_genes_from_organotypic_markers.py` - `Figure4_5/02_Pseudobulk_network/04_EC_organotypic_marker.ipynb` - `Figure4_5/02_Pseudobulk_network/05_pseudobulk_network_input.ipynb` - `Figure4_5/02_Pseudobulk_network/06_piechart_network_Figure5.ipynb` - `Figure4_5/02_Pseudobulk_network/07_Supplementary_Figure10_C.ipynb` - `Figure4_5/03_mouse_vs_human/Figure4_D_&_Supplementary_Figure10_A.ipynb` - `Figure4_5/03_mouse_vs_human/Supplementary_Figure8_C_D.ipynb` - `Figure4_5/03_mouse_vs_human/Supplemnetary_Figure10_B.ipynb` Main input objects: - `h5ad/Fig4_h5ad/for_pseudobulk_organotypic_onlycapi.h5ad` - `h5ad/Fig4_h5ad/Fig4_capiilary_human.h5ad` - `h5ad/Fig4_h5ad/Fig4_All_celltyp_human.h5ad` - `h5ad/Fig3_h5ad/09_subcluster_capi_harmony_commo_speci.h5ad` (as the mouse capillary) Representative analyses: - Figure 4A-B: organ-specific capillary DEG sets and Jaccard similarity clustering, - organotypic pseudobulk differential expression analysis across endocrine organs, followed by refinement of organ-specific markers through removal of non-endothelial-enriched genes identified by conserved-marker pseudobulk analysis, - cross-species comparison of human and mouse capillary gene expression based on DEGs identified by pseudobulk differential expression analysis, - Figure 5A-E: network construction using significant terms identified by hypergeometric testing as nodes and filtered Jaccard index values representing gene overlap between terms as edges, - topological analysis of the resulting network using betweenness centrality and within-module z-scores to identify important nodes. Key parameters visible in `01_EC_organotypic_run_pseudobulk_DEG.py`: - organ list: `Thyroid`, `Pituitary-Gland`, `Adrenal-Gland`, `Brain`, `Liver`, `Kidney`, - pseudo-replicates are randomly assigned into `10` per organ, - pseudobulk counts are generated from `adata.layers["counts"]`, - differential expression is performed with `pydeseq2` using design factor `focus`. ### Figure 6. Inflammatory regulons, GSVA/pseudobulk pathway summaries, and organ-level cell-cell interaction analysis Representative notebooks/scripts: - `Figure6/01_scenic/Figure6_A_&_Supplementary_Figure12.ipynb` - `Figure6/01_scenic/Figure6_B.ipynb` - `Figure6/01_scenic/Supplementary_Figure12.ipynb` - `Figure6/02_GSVA/01_Figure6C_Supp_Fig13.ipynb` - `Figure6/03_CCI/01_CCI_per_organ.py` - `Figure6/03_CCI/02_Supplementary_Figure14.ipynb` - `Figure6/03_CCI/03_1_make_interaction_counts.py` - `Figure6/03_CCI/03_2_Figure6_E.ipynb` - `Figure6/03_CCI/04_Figure6_D.ipynb` - `Figure6/03_CCI/05_run_CCI_70%sampling_repeat.py` - `Figure6/03_CCI/05_Supplementary_Figure14_B.ipynb` Main input objects: - `h5ad/FIg2_h5ad/subcluster_EC_harmony_Anno.h5ad` - `h5ad/03_SNU_cellbender_filtered_QC.h5ad` - `h5ad/SNU_cellbender_filtered_BBKNN_Anno.h5ad` Representative analyses: - organ-specific AUCell score analysis of inflammation-related transcription factors using regulon analysis, - pseudobulk GSVA-based pathway activity scoring for inflammatory gene sets, - analysis of organ-specific cell-cell interaction (CCI) patterns, - comparison of the strength of inflammation-related ligand-receptor pairs across organs, - interaction count matrix generation and robustness analysis using repeated 70% subsampling. Key parameters visible in `Figure6/03_CCI/01_CCI_per_organ.py`: ### Figure 7. Thyroid aging/PTC endothelial analysis, scoring, communication, and survival Representative preprocessing notebooks/scripts inside Figure 7: - `Figure7/Thyroid_Normal_PTC/Preprocess/01_QC/01_run_scrublet.py` - `Figure7/Thyroid_Normal_PTC/Preprocess/01_QC/02_run_QC_concat.py` - `Figure7/Thyroid_Normal_PTC/Preprocess/02_Batch_corr/03_1_BBKNN.py` - `Figure7/Thyroid_Normal_PTC/Preprocess/02_Batch_corr/03_1_low_annotation.ipynb` - `Figure7/Thyroid_Normal_PTC/Preprocess/02_Batch_corr/03_2_sub_endo_harmony.ipynb` - `Figure7/Thyroid_Normal_PTC/Preprocess/02_Batch_corr/03_3_sub_endo_harmony_annotation.ipynb` Representative downstream notebooks/scripts: - `Figure7/Thyroid_Normal_PTC/05_DEG_hypergeo/Figure7_B.ipynb` - `Figure7/Thyroid_Normal_PTC/07_Scoring/Figure7_C.ipynb` - `Figure7/Thyroid_Normal_PTC/06_CCI/run_CCI_batch.py` - `Figure7/Thyroid_Normal_PTC/06_CCI/Figure7_D.ipynb` - `Figure7/Thyroid_Normal_PTC/06_CCI/Figure7_E.ipynb` - `Figure7/Thyroid_Normal_PTC/06_CCI/Figure7_F.ipynb` - `Figure7/Thyroid_Normal_PTC/06_CCI/Figure7_G.ipynb` - `Figure7/Thyroid_Normal_PTC/06_CCI/Supplemnetary_Figure17.ipynb` - `Figure7/Thyroid_Normal_PTC/09_Survival/TCGA_PTC_survival_run.py` Main input objects: - `h5ad/Fig7_h5ad/FIg7_Thyroid_BBKNN_hvg3500_low_anno.h5ad` - `h5ad/Fig7_h5ad/FIg7_Thyroid_EC_subclustering_anno.h5ad` - `h5ad/Fig7_h5ad/for_cci/young.h5ad` - `h5ad/Fig7_h5ad/for_cci/old.h5ad` - `h5ad/Fig7_h5ad/for_cci/ptc.h5ad` Representative analyses: - thyroid sample QC, integration, low-level annotation, and endothelial-focused subclustering, - hypergeometric test of condition-associated DEG sets for Figure 7B, - endothelial hallmark/pathway scoring in Figure 7C using gene-set files in `Figure7/Thyroid_Normal_PTC/07_Scoring/geneset/`, - CCI analysis in each of the young, old, and ptc thyroid objects, - interaction strength of inflammation-related ligand-receptor pairs across conditions in Figure 7E-G - TCGA-PTC survival analysis after stratifying samples into high and low groups based on the average expression of the ligand and receptor genes.