04-feature-selection-Z-distance.Rmd

```{r parameters-and-defaults, include = FALSE}
module <- "scRNAseq"
section <- "feature_selection"
```

```{r parameter-merge, include = FALSE}
local_params <- module %>%
  options() %>%
  magrittr::extract2(module) %>%
  magrittr::extract2(section) %>%
  ReporteR.base::validate_params(parameters_and_defaults)
```

```{r scRNAseq-feature-selection-Z-distance-checks}
assertive.sets::assert_is_subset(local_params$assay, SummarizedExperiment::assayNames(object_filtered))
```

### Transcriptional noise within known celltypes

In total, **`r sum(SummarizedExperiment::rowData(object_filtered)$is_het)` heterogeneous genes** have been detected using methods described above. Heterogeneous genes can be used to calculate and compare the overall *transcriptional noise* in groups of cells, e.g. cell types, cell clusters, etc. Cell-to-cell distance values (here calculated as 1 - `r local_params$wgcna_cor_method` correlation) serve as a good proxy for transcriptional noise within and between cell populations, as they characterize cells in a defined state (low noise, low distance) *vs.* cells that loose their characteristic pattern (high noise, high distance) [@mojtahedi_fate_2016]. Figure \@ref(fig:scRNAseq-feature-selection-Z-distance-figure) shows transcriptional noise within and between cells across the experimental factor `r local_params$celltype`.

```{r scRNAseq-feature-selection-Z-distance-processing, include = FALSE}
celltypes <- droplevels(factor(SummarizedExperiment::colData(object_filtered)[, local_params$celltype]))
data <- SummarizedExperiment::assay(object_filtered, local_params$assay)[SummarizedExperiment::rowData(object_filtered)$is_het,]

transcriptome_noise_list <- lapply(levels(celltypes), function(c) {
  i <- which(celltypes == c)
  cor <- WGCNA::cor(data[, i], method = local_params$wgcna_cor_method)
  1 - cor[upper.tri(cor)]
})

transcriptome_noise <- data.frame(celltype = rep(levels(celltypes), times = sapply(transcriptome_noise_list, length)), distance = unlist(transcriptome_noise_list))
```

```{r scRNAseq-feature-selection-Z-distance-figure, echo = FALSE, message=FALSE, warning=FALSE, fig.cap = "Cell-to-cell distance as a proxy of transcriptional noise. Low distance values hint to low transcriptional noise within a homogeneous cell population of a defined state. High distance values indicate cells that are loosing their characteristic state."}
theme_violins <- ggplot2::theme(plot.background = ggplot2::element_blank(),
                                panel.grid.major = ggplot2::element_blank(),
                                panel.grid.minor = ggplot2::element_blank(),
                                panel.border = ggplot2::element_blank(),
                                panel.background = ggplot2::element_blank(),
                                axis.line.y = ggplot2::element_line(size=.3),
                                axis.line.x = ggplot2::element_line(size=.3),
                                legend.text = ggplot2::element_text(size=6),
                                axis.title.y = ggplot2::element_text(color="black", size=10),
                                axis.title.x = ggplot2::element_blank(),
                                plot.title = ggplot2::element_text(face="bold", color="black", size=6),
                                legend.key.size =  grid::unit(2, "mm"),
                                legend.margin = grid::unit(-25, "mm"),
                                strip.background = ggplot2::element_blank(),
                                strip.text = ggplot2::element_blank(),
                                axis.text = ggplot2::element_text(color="black", size=8),
                                axis.text.x = ggplot2::element_text(color="black", size=10))

ggplot2::ggplot(data = transcriptome_noise, ggplot2::aes(x = celltype, y = distance, fill = celltype)) +
  ggplot2::geom_violin(trim = F) +
  ggplot2::geom_boxplot(width = 0.1, fill = "white") +
  #scale_fill_manual(values = nkx_colors) +
  ggplot2::scale_y_continuous(expand = c(0.05, 0)) +
  ggplot2::guides(fill = FALSE, colour = FALSE) +
  ggplot2::xlab("") + ggplot2::ylab("Cell to cell distance") + theme_violins +
  ggplot2::theme(plot.margin = grid::unit(c(0, -5, 2.5, 0), "mm"))
```