05-dimension-reduction-B-tsne-dist.Rmd

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

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

```{r scRNAseq-dimension-reduction-B-tsne-dist-checks, include = FALSE}
assertive.sets::assert_is_subset(local_params$features, colnames(SummarizedExperiment::colData(object_filtered)))
```

### t-SNE based on cell-to-cell correlation

**t**-**S**tochastic **N**eighbor **E**mbedding[@maaten_tsne_2008] (t-SNE) converts a high-dimensional data set into a low-dimensional representation using pairwise similarity between objects of the data set. t-SNE claims to be able to capture much of the local structure in a data set, while also revealing global structures like *clusters*. t-SNE adapts to the data by applying different transformations on different regions in the data space, using many iterations over the data. Further, t-SNE allows to balance the importance of local vs. global structures via an important *perplexity* parameter, which has, together with the data adaptation step, complex effects on the resulting dimensionality reduction. 

```{r scRNAseq-dimension-reduction-B-tsne-dist-processing, include=FALSE, echo=FALSE}
input <- SummarizedExperiment::assay(object_filtered, local_params$assay)

if (local_params$subset_het) {
  input <- input[SummarizedExperiment::rowData(object_filtered)$is_het, ]
}

dissimilarity <- (1 - WGCNA::cor(input, method = local_params$wgcna_cor_method)) / local_params$wgcna_cor_denominator

if (local_params$tsne_dist) {
  di <- as.dist(dissimilarity)
} else {
  di <- dissimilarity
}

# Allow choosing perplexity automatically
if (local_params$tsne_perplexity < 0) {
  perplexity <- floor(ncol(as.matrix(di))/5)
} else {
  perplexity <- local_params$tsne_perplexity
}

tsne <- Rtsne::Rtsne(as.matrix(di), dims = local_params$dims, is_distance = local_params$tsne_dist, perplexity = perplexity, max_iter = local_params$maxiter)
SingleCellExperiment::reducedDim(object_filtered, "tsne-dist") <- tsne$Y
```

```{r scRNAseq-dimension-reduction-B-tsne-dist-figure-params, include = FALSE, echo = FALSE}
fig_height <- ReporteR.base::estimate_figure_height(
  height_in_panels = 1,
  panel_height_in_in = params$formatting_defaults$figures$panel_height_in,
  axis_space_in_in = params$formatting_defaults$figures$axis_space_in,
  mpf_row_space = as.numeric(grid::convertUnit(grid::unit(5, 'mm'), 'in')),
  max_height_in_in = params$formatting_defaults$figures$max_height_in)


caption_tsne <- glue::glue("Dimensionality reduction using t-SNE [@maaten_tsne_2008] on a dissimilarity structure (calculated as *1 - {local_params$wgcna_cor_method} correlation*) with a perplexity value of {perplexity} and {local_params$tsne_iter} epochs of learning.")
```

```{r scRNAseq-dimension-reduction-B-tsne-dist-figure, echo = FALSE, message=FALSE, warning=FALSE, fig.height = fig_height$global, fig.cap = caption_tsne}
figure_dimred_tsne_dist <- multipanelfigure::multi_panel_figure(height = fig_height$sub, columns = min(3, length(local_params$features)), rows = 1, unit = "in")

for(i in 1:min(3, length(local_params$features))) {
  tmp_plot <- object_filtered %>%
    scater::plotReducedDim(use_dimred = "tsne-dist", colour_by = local_params$features[i], add_ticks = FALSE) +
    ggplot2::guides(colour = FALSE) +
    theme_dimred_scatter
  figure_dimred_tsne_dist <- multipanelfigure::fill_panel(figure_dimred_tsne_dist, tmp_plot)
}

figure_dimred_tsne_dist
```