update detailed example

vignettes/detailed_example.Rmd

@@ -16,6 +16,7 @@ library(dplyr)
 library(ggplot2)
 library(tidyr)
 library(immunedeconv)
+library(tibble)
 ```
 
 *Immunedeconv* ships with an example dataset with samples from four patients with metastatic melanoma published in @EPIC2017. 
@@ -60,9 +61,11 @@ res_quantiseq %>%
 
 We observe that
 
-* the scores of the first two samples sum to 1, while the last two don't. #TODO other cells
-* the first two samples contain a large quantity of B cells and CD4+ T cells. 
+* the first two samples (LAU355, LAU1314) appear to contain a large amount of CD4+ T cells and B cells
+* the last two samples (LAU1255, LAU125) appear to contain a large amount of "uncharacterized cells"
+* the last sample (LAU125) appears to contain no CD8+ T cells, often associated with bad prognosis. 
 
+Estimating the amount of "uncharacterized cells" is a novel feature introduced by quanTIseq and EPIC [@EPIC2017, @quantiseq2017]. This estimate often corresponds to the fraction of tumor cells in the sample. 
 
 
 Let's now apply MCP-counter to the same dataset.
@@ -85,7 +88,43 @@ res_mcp_counter %>%
     theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))
 ```
 
-With the scores being in arbitrary units, the results are not useful for judging if a cell type is present in the sample, or not. However, the score correctly reflects the presence of NK cells and CD8+ T cells in the corresponding samples.
+With the scores being in arbitrary units, the results are not useful for judging if a cell type is present in the sample, or not. However, we can compare the relative values between samples and relate them to the results we obtained earlier using quanTIseq. 
 
+* Consistent with quanTIseq, MCP-counter predicts B cells to be more abundant in LAU355 and LAU1314 than in the other two samples. 
+* Also consistent with quanTIseq, the last LAU125 appears to contain significantly less CD8+ T cells than the other samples. 
+* the estimates for Neutrophils and Monocytes are inconsistent with quanTIseq. 
 
-## TODO add comparison with FACS data. 
+
+## Comparison with FACS data
+Let's now compare the results with 'gold standard' FACS data obtained for the four samples. This is, of course, not a representative benchmark, but it gives a notion about what magnitude of predictive accuracy we can expect. 
+```{r}
+# construct a single dataframe containing all data
+#
+# re-map the cell-types to common names. 
+# only include the cell-types that are measured using FACS
+cell_types = c("B cell", "T cell CD4+", "T cell CD8+", "NK cell")
+
+tmp_quantiseq = res_quantiseq %>% map_result_to_celltypes(cell_types, "quantiseq") %>% 
+  rownames_to_column("cell_type") %>%
+  gather("sample", "estimate", -cell_type) %>% 
+  mutate(method="quanTIseq") 
+tmp_mcp_counter = res_mcp_counter %>% map_result_to_celltypes(cell_types, "mcp_counter") %>%
+  rownames_to_column("cell_type") %>%
+  gather("sample", "estimate", -cell_type) %>%
+  mutate(method="MCP-counter") 
+
+result = bind_rows(tmp_quantiseq, tmp_mcp_counter) %>%
+  inner_join(dataset_racle$ref)
+```
+
+Plot the true vs. estimated values: 
+```{r, fig.width=6, fig.height=7}
+result %>% 
+  ggplot(aes(x=true_fraction, y=estimate)) + 
+    geom_point(aes(shape=cell_type, color=cell_type)) + 
+    facet_wrap(cell_type~method, scales="free_y", ncol = 2) + 
+    scale_color_brewer(palette = "Dark2") + 
+    theme_bw()
+```
+
+(MCP counter does not provide estimates for CD4+ T cells.)

vignettes/immunedeconv.Rmd

@@ -196,6 +196,7 @@ The algorithm is explained in detail in the methods section of @sturm2019.
 In general, cell-type scores allow for the comparison (1) between samples, (2) between cell-types or (3) both. 
 Between-sample comparisons allow to make statements such as *"In patient A, there are more CD8+ T cells than in patient B"*. 
 Between-cell-type comparisons allow to make statements such as *"In a certain patient, there are more B cells than T cells"*.
+For more information, see our Benchmark paper (@sturm2019). 
 
 ### Methods that allow between-sample comparisons