Harmonise the datasets

[stemangiola]

we need to

1) integrate the raw counts to 2) cluster the data and 3) use the existing annotation to label each cluster.

• I will provide you with this pipeline so you can try to execute it on the WEHI servers. This pipeline takes care of dead cells and doublets and does SCT normalisation with no factors
• Take each Seurat file, run clustering and plot ribosomal score and mitochondrial score (they will be already ion the datasets), and see if you get cluster based on those scores, then you re normalise based on those factors
• you will use the output being SCT normalised dataset to integrate at the sample level

Cell annotation

• [ ] Re-annotate the Seurat cluster (resolution 0.6).

• [ ] Look at each single cluster and look at cell types. If the cell types in one cluster do not agree with each other; increase the resolution (discuss with Stefano).

scSHC

• [x] Pick max 500 cells from each sample, run findCluster with the integrated assay.

[XpelC]

1. integrate the raw counts to

   The integrated data group by dataset.

   

The integrated data group by sample.

The integrated data split by dataset

The integrated data group by cell types (before relabeling)

[stemangiola]

• [ ] test different number of clusters (changing resolution parameter), e.g. 10, 15, 20, 25, 30

https://github.com/igrabski/sc-SHC

HERE WE MEET AGAIN WITH A REPORT OF ALL CLUSTERING TO CHOOSE THE BEST

• [ ] make summary of cell count per dataset per cluster

cluster | dataset_1 | dataset_2   | ...
1           | 200       | 500                 | ...
12         | 0          | 1200                 | ...

• [x] For each cluster, how many cell types

cluster    | cxell_type 
1             | t_cell           
1             | t_CD8_memory           
1             | t_CD4_exhausted
2             | monocytes

• [ ] After you build the list of cell-type by cluster, you try to label the cluster yourself finding a consensus across the different cell types of the different studies

for example

cluster 1 includes 5 datasets and includes, cd8_memory_t_cell, cd8_memory_T_cell, t_exhausted, cd8_resident. So you could assume this cluster cd8_memory_t_cell_exhausted

x_y_z, x_y > x_y_z because x_y is also present in another cluster, so the z flavour could distinguish this particular cluster.

[XpelC]

• For each cluster, how many cell types

1. umap result (cluster resolution 2, pca 50 dimensions)

line seurat_clusters cell_types

1 0 CAFs,Cancer,Cancer/Epithelial,Endothelial,Epithelial cell,Epitheli… 2 1 CD4 naive-cm,CD4 Trm,CD8 cytotoxic,CD8 Trm,CD8+ Tem,Doublets,Mast,… 3 2 Basal/intermediate,Cancer,CE,Endothelial,Endothelial 1,Endothelial… 4 3 Cancer,Cancer/Epithelial,CE,Doublets,Endothelial,Epithelial cell,E… 5 4 Cancer,Cancer/Epithelial,CE,Endothelial,Epithelial cell,Epithelial… 6 5 CD4 naive-cm,CD4 Trm,CD8 cytotoxic,CD8 Trm,CD8+ Tem,Mast,Monocytic… 7 6 Basal/intermediate,BE,Cancer,Cancer/Epithelial,CE,Epithelial cell,… 8 7 CAFs,Cancer,Cancer/Epithelial,CE,Doublets,Endothelial,Epithelial c… 9 8 Basal/intermediate,Cancer,Cancer/Epithelial,CE,Doublets,Endothelia… 10 9 CD4 naive-cm,CD4 Trm,CD8 Trm,CD8+ Tem,Doublets,Epithelial cell,Epi… 11 10 Basal/intermediate,Cancer,Cancer/Epithelial,CE,Epithelial cell,Epi… 12 11 CAFs,Cancer,Cancer/Epithelial,Epithelial cell,Epithelial_cells,Fib… 13 12 BE,Cancer,Cancer/Epithelial,CE,Epithelial cell,Epithelial_cells,LE… 14 13 Apidocytes,CAFs,Cancer,Cancer/Epithelial,Endothelial,Epithelial ce… 15 14 CAFs,Cancer,Cancer/Epithelial,CE,Doublets,Endothelial,Epithelial c… 16 15 Basal/intermediate,Cancer,Cancer/Epithelial,CE,Epithelial cell,Epi… 17 16 DC,Endothelial,Mac1,Macrophage,Mast,Mono,Monocyte,Monocytes,Monocy… 18 17 Basal/intermediate,Cancer,Cancer/Epithelial,Epithelial cell,Epithe… 19 18 Basal/intermediate,Cancer,Cancer/Epithelial,CE,Epithelial cell,Epi… 20 19 Basal/intermediate,Cancer,Cancer/Epithelial,CE,Epithelial cell,Epi… 21 20 Cancer,Cancer/Epithelial,Epithelial cell,Epithelial_cells,LE-KLK3,… 22 21 Endothelial,Endothelial cell,Endothelial_cells,Luminal,Unassigned 23 22 CD8 Trm,CD8+ Tem,Epithelial cell,Epithelial_cells,LE-KLK4,Luminal,… 24 23 DC,Endothelial,Luminal,Macrophage,Mast,Monocyte,Monocytes/Macropha… 25 24 Cancer,Cancer/Epithelial,CE,Epithelial cell,Epithelial_cells,LE-KL… 26 25 Cancer,Cancer/Epithelial,Cancer/Epithelial Cycling,Epithelial cell… 27 26 Apidocytes,CAFs,Cancer,Endothelial,Fibroblast,Myofibroblast,PVL ce… 28 27 B cell,Cancer,Cancer/Epithelial,Epithelial cell,Epithelial_cells,L… 29 28 Cancer,Cancer/Epithelial,Epithelial cell,Epithelial_cells,LE-KLK3,… 30 29 Endothelial,Endothelial cell,Endothelial_cells,Epithelial_cells,Fi… 31 30 B cell,B-cells,Luminal,Mast,Monocyte,Monocytes/Macrophages,T 32 31 CD8+ Tem,Luminal,NK_cell,T,T cell,T-cells 33 32 Cancer,Cancer/Epithelial,Epithelial cell,Epithelial_cells,LE-KLK3,… 34 33 Endothelial,Macrophage,Monocyte,Monocytes/Macrophages,Monocytic,My… 35 34 T cell,T-cells 36 35 Cancer,Cancer/Epithelial,Epithelial cell,Epithelial_cells,Fibrobla… 37 36 Monocytes,Monocytes/Macrophages,Myeloid,Pre-B_cell_CD34- 38 37 Epithelial cell,Luminal,Monocytes/Macrophages 39 38 Cancer/Epithelial,Luminal 40 39 pDCs

[stemangiola]

Much better!

would you be able to annotate this atlas? One cell type per cluster?

[XpelC]

Hello Stefano,

I’ll try it tomorrow, and test if the samples within the dataset need integration by plotting umap.

Best wishes, Xinpu

On Sep 11, 2022, at 6:13 PM, Stefano Mangiola @.**@.>> wrote:

Much better!

would you be able to annotate this atlas? One cell type per cluster?

— Reply to this email directly, view it on GitHubhttps://github.com/stemangiola/cellsig/issues/64#issuecomment-1242913765, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ALBA44YNA4KWZ4AU2UL6DILV5WIBJANCNFSM54BXBJNA. You are receiving this because you commented.Message ID: @.***>

[stemangiola]

I’ll try it tomorrow, and test if the samples within the dataset need integration by plotting umap.

Good. remember to do things in parallel. One of these days you can present to the consortium meeting if you wish. Presenting your work.

[XpelC]

1. umap result (cluster resolution 2, pca 50 dimensions)

Try 100 pcs and plot the table.

[stemangiola]

@XpelC ,

this is a function to plot heatmap of the markers of just one cluster (with a balance sampling of cells to zoom onto that cluster)

gamma_delta_plot = 
    data_lymphoid %>% 
    get_markers_one_vs_all(
        curated_cell_type_pretty %in% c("gamma_delta"), 
        cluster_col = "curated_cell_type_pretty", 
        assay = "SCT", 
        disp.min = 0, disp.max = 4, size=3
    ) +
    viridis::scale_fill_viridis( option = "A")

source the attached file before

functions.R.zip

[XpelC]

• For each cluster, how many cell types

(With breast cancer filtered)

A tibble: 45 × 2

seurat_clusters cell_types

1 0 cancer,epithelial,luminal,T 2 1 cancer,CD8_Tem,endothelial,epithelial,luminal,NK,perivascular,smoo… 3 2 B,cancer,epithelial,luminal,mast,T 4 3 cancer,CD8_Tem,endothelial,epithelial,fibroblast,luminal,mast,T 5 4 B,basal,basal_intermediate,cancer,CD4_naive,CD8_Tem,club_cell,endo… 6 5 cancer,CD8_Tem,CD8_Trm,club_cell,epithelial,luminal,mast,NK,periva… 7 6 basal,basal_intermediate,cancer,club_cell,endothelial,epithelial,f… 8 7 dendritic,endothelial,luminal,mac,macrophage,mast,monocyte,myeloid… 9 8 B,CD4_naive,CD8_Tem,CD8_Trm,epithelial,luminal,monocyte,NK,T 10 9 cancer,epithelial,luminal 11 10 cancer,epithelial,luminal 12 11 B,CD4_naive,CD4_Trm,CD8_cytotoxic,CD8_Tem,CD8_Trm,endothelial,mast… 13 12 cancer,CD8_Tem,endothelial,epithelial,luminal,mast,T 14 13 cancer,epithelial,luminal,T 15 14 basal,basal_intermediate,club_cell,endothelial,epithelial,fibrobla… 16 15 epithelial,luminal,perivascular,T 17 16 endothelial,fibroblast,mast 18 17 apidocytes,cancer_associated_fibroblast,endothelial,fibroblast,lum… 19 18 B,cancer,epithelial,luminal,plasma,T 20 19 cancer,epithelial,luminal 21 20 cancer,epithelial,luminal,NK,perivascular,sperm,T 22 21 cancer,epithelial,luminal 23 22 endothelial,mast 24 23 cancer,epithelial,luminal 25 24 epithelial,luminal,monocyte,sperm,T 26 25 apidocytes,cancer_associated_fibroblast,endothelial,fibroblast,myo… 27 26 endothelial,epithelial 28 27 epithelial,luminal,mast,perivascular,T 29 28 basal_intermediate,cancer,club_cell,epithelial,hillock,luminal,mas… 30 29 apidocytes,club_cell,endothelial,epithelial,fibroblast,monocyte,my… 31 30 cancer,epithelial,luminal 32 31 cancer,endothelial,epithelial,luminal,mast 33 32 epithelial,luminal,T 34 33 epithelial,luminal 35 34 luminal,mast 36 35 T 37 36 epithelial,mast 38 37 fibroblast 39 38 CD8_Tem,epithelial,monocyte,NK,T 40 39 epithelial,luminal 41 40 B 42 41 CD8_Trm,monocyte 43 42 mast 44 43 epithelial 45 44 luminal

[stemangiola]

Good, you can now use DoHeatmap from Seurat to annotate and check the clusters with their marker genes.

[XpelC]

Do you want me to divide them into two groups first? (Epithelial and immune?)

On Sep 19, 2022, at 9:43 AM, Stefano Mangiola @.**@.>> wrote:

Good, you can now use DoHeatmap from Seurat to annotate and check the clusters with their marker genes.

— Reply to this email directly, view it on GitHubhttps://github.com/stemangiola/cellsig/issues/64#issuecomment-1250412714, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ALBA444T3ZMOLZBQIUOCKZDV66SIZANCNFSM54BXBJNA. You are receiving this because you were mentioned.Message ID: @.***>

[stemangiola]

Hello @XpelC , we need to proceed fast.

1) Please address the macro clusters, the epithelial labelling is obviously including wrond microclusters. Please see and address this issue

https://github.com/stemangiola/cellsig/issues/69#issuecomment-1274126773

2) we mention for the heatmap:

• define the macrocluster (issue 1 above), please use care in your work, I see some mistakes that can be avoided.
• maker an heatmap within each macrocluster

I would expect points 1 and 2 done in one day. Please send me the fix of point one in the other github issue first to wait for my confirmation and then you will proceed with heatmapping.