Project 1
Tumor cells form complex ecosystems with other stromal cells. These ecosystems influence tumor initiation and progression as well as response to treatment. Single-cell and spatial transcriptomics data allows to investigate the interactions between cell types by leveraging information about receptor and ligand gene expression.
In the context of the lungNENomics project, we have performed spatial transcriptomics (VISIUM) on 4 "supra-carcinoid" samples, the most aggressive molecular subtype of lung neuroendocrine tumors resembling the high-grade and deadly large-cell neuroendocrine carcinomas although they have the morphology of lower-grade carcinoids (Alcala et al. Nat Commun 2019, see Figure below ). Bulk sequencing analyses suggest that supra-carcinoids experience a very specific microenvironmental niche, with many immune cells. The project consists in investigating the cellular interactions between tumor and its microenvironment.
Fig. 1 | Low-dimensional representation of lung neuroendocrine neoplasms from the lungNENomics cohort (multi-omic factor analysis). From Alcala et al. Nat Commun 2019.
- Processed VISIUM data (R object) for 4 samples, available on osiris in /data/Training-MG/files/data/Project1_spatial_lnets/data/adatas_scanpy_copykat_06082024.Rdata (annotated data) and IRIS_object.rdata (IRIS object containing sptaial domains) . Caution, data is read only and should stay on the server (no copy).
Scripting in R, data exploration, statistics
- Install the cell-cell interaction R package CellChat v2 (see dependencies and instructions at https://github.com/jinworks/CellChat)
- Load the data for the 4 tumors, prepare the input for cellchat v2
- Run the method following the tutorial for analyzing multiple spatial samples (https://htmlpreview.github.io/?https://github.com/jinworks/CellChat/blob/master/tutorial/CellChat_analysis_of_multiple_spatial_transcriptomics_datasets.html); see also procedure 1 for single-cell analysis from the paper for detailed explanations about some steps (https://www.nature.com/articles/s41596-024-01045-4#Sec22). Note : first use the spatial domains in the IRIS_object file as labels for the spots. You might explore using the cell types obtained with spot deconvolution (in adatas_scanpy_copykat_06082024) as a second step, assigning as a label the most prevalent cell type in each spot
- Visualise and interpret the results. Are there domains with particularly complex interactions?
- Discuss limitations/biases in the study and importance for these aggressive lung neuroendocrine tumors.
- Conceptual understanding of the subject (reading the papers to understand the method is essential)
- R scripting