PURPOSE Immune response to tumors defines patient outcomes and our goal was to identify where tumors lie on the self/nonself axis and if they exploit selfness to subvert immune destruction. MATERIALS AND METHODS A new measure for quantifying selfness was developed using transcriptomes based on a selfness model stemming from thymic education. The thymus-likeness score (TLS) was validated on data from the Human Protein Atlas and then used to interrogate tumor recognizability across 32 solid tumor cohorts from The Cancer Genome Atlas (TCGA). A robust statistical pipeline was devised to identify links between the TLS and disease outcomes such as immune infiltration and patient survival. RESULTS We first developed a novel selfness score (TLS) for estimating the T-cell recognizability in any sample, given its transcriptome. Using the TLS, we interrogated tumor recognizability across 32 solid tumor cohorts from TCGA and found that there is a consistent increase in selfness in tumors compared with their adjacent nontumor tissues. In our pan-cancer analysis, we found an unexpected inverse association between TLS and immune infiltration of macrophages, dendritic cells, CD8 T cells, and Treg cells. We found the selfness, when combined with the immune infiltration, to correlate with patient survival in multiple cohorts. CONCLUSION The TLS is one of the first models for quantifying tumor recognizability from an immune perspective. The increase in selfness in tumors, along with its inverse relationship with immune infiltration, and links to survival hints at the presence of an immune sculpting selection mechanism that selects for two types of tumors: (1) easily recognizable, but poorly infiltrated and (2) heavily infiltrated, but poorly recognizable. Our results provide a new direction for investigating tumor evolution trajectories.
Sakthivel et al. (Fri,) studied this question.