Table 2.
Guidelines for method selection
| Cell type | Recommended methods | Overall performance | Absolute score | No background predictions |
|---|---|---|---|---|
| B cell | EPIC | ++ | ++ | + |
| MCP-counter | ++ | − | − | |
| T cell CD4+ | EPIC | ++ | ++ | − |
| xCell | ++ | − | ++ | |
| T cell CD4+ non-regulatory | quanTIseq | + | ++ | + |
| xCell | + | − | ++ | |
| T cell regulatory | quanTIseq | ++ | ++ | − |
| xCell | ++ | − | ++ | |
| T cell CD8+ | quanTIseq | ++ | ++ | − |
| EPIC | ++ | ++ | − | |
| MCP-counter | ++ | − | − | |
| xCell | + | − | ++ | |
| Natural Killer Cell | EPIC | ++ | ++ | + |
| MCP-counter | ++ | − | − | |
| Macrophage / Monocyte | xCell | − | ++ | |
| EPIC | + | ++ | + | |
| MCP-counter | ++ | − | − | |
| Cancer-associated fibroblast | EPIC | ++ | ++ | + |
| MCP-counter | ++ | − | − | |
| Endothelial Cell | EPIC | ++ | ++ | + |
| xCell | ++ | − | ++ | |
| Dentricic cell | None of the methods can be recommended to estimate overall DC content. MCP-counter and quanTIseq can be used to profile mDCs. | |||
Note: For each cell type, we recommend which method to use, highlighting advantages and possible limitations. Overall performance: indicates how well predicted fractions correspond to known fractions in the benchmark. Absolute score: the method provides an absolute score that can be interpreted as a cell fraction. Methods that do not provide an absolute score only support inter-sample comparisons within the same experimental dataset, i.e. the score is only meaningful in relation to another sample. Background predictions: indicates, if a method tends to predict a cell type although it is absent.