What T cells are dysfunctional, actually - and how to find out?

[grst]

In the Li 2018, Cell publication they find a continuum between cytotoxic cells and cell expression exhaustion markers. They also show that cells expression markers of what has previously been called "exhaustion" are actually highly preoperative.

In my data, on first sight, cytotoxic markers such as Granzyme B co-localize with LAG3 and TIM3 in the umap plot.

This rises the question which cells are actually dysfunctional. While this is probably beyond of the scope of the thesis, I am wondering if there is a possibility to do an unbiased way (i..e. independent of known marker genes) to define functional states of T cells.

First idea:

• unsupervised clustering of all T cells
• differential expression analysis for each cluster
• compute enrichment score across TCGA
• associate with survival data

[grst]

Second idea:

• Compute effect on survival for each gene (expression vs. survival). Note effect size for each gene
• For each cell, compute a "survival score", something like "sum of (gene expression * effect size for each gene)"
• -> hopefully cells with a high score co-localize

[FFinotello]

This paper provides curated survival data for TCGA cancers and recommendations on the endpoints to use for each cancer type: https://www.cell.com/cell/fulltext/S0092-8674(18)30229-0.