02-quality-control-C-tsne.Rmd

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

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

```{r scRNAseq-quality-control-C-tsne-checks}
assertive.sets::assert_is_subset(local_params$features, colnames(SummarizedExperiment::colData(object)))
```

```{r scRNAseq-quality-control-C-tsne-processing, include = FALSE}
object %<>%
  singlecellutils::calculate_qc_map(features = local_params$features, perplexity = 51, max_iter = 2000)
```

### t-SNE based quality map
Here we use **t**-**S**tochastic **N**eighbor **E**mbedding[@maaten_tsne_2008] (t-SNE), an algorithm for unsupervised dimension reduction, to visualize single cells in a *sample quality subspace*. t-SNE works on the assumption that a low-dimensional structure exists, although objects are embedded in a high dimensional space. In case of the quality subspace, samples are embedded in a high dimensional space described by the features `r ReporteR.base::itemize(local_params$features)`. Typically, t-SNE arranges samples that are associated with low quality features in close proximity and enables exclusion of those samples from downstream analysis (Figure \@ref(fig:scRNAseq-quality-control-C-tsne-figure)).

```{r scRNAseq-quality-control-C-tsne-figure, warning = FALSE, message = FALSE, echo = FALSE, fig.cap = "Biplot of dimensions 1 vs. 2 from *t-SNE* [@maaten_tsne_2008] of the sample quality subspace. Percentages in the axis labels indicate the fractional variance explained by the corresponding dimension."}
object %>%
  scater::plotReducedDim(use_dimred = "qcmap", colour_by = local_params$features[1]) +
  ggplot2::guides(colour = FALSE) +
  theme_qc_pca
```