Add captions for figures 3 & 4

content/100.figure-table-legends.md

@@ -38,3 +38,39 @@ Points are colored by whether the cell was kept or removed, as determined by bot
 F. UMAP embeddings of log-normalized RNA expression values where each cell is colored by the number of genes detected.
 G. UMAP embeddings of log-normalized RNA expression values for the top four most variable genes, colored by the given gene's expression.
 In the actual summary QC report, the top 12 most highly variable genes are shown.
+
+
+![**Figure 3. ScPCA Portal project download file structure and merged object workflow.**](https://raw.githubusercontent.com/AlexsLemonade/scpca-paper-figures/main/figures/compiled_figures/pngs/figure_3.png?sanitize=true){#fig:fig3 width="7in"}
+
+A. File download structure for an ScPCA Portal project download in `SCE` format.
+The download folder is named according to both the project ID and the date it was downloaded.
+Download folders contain one folder for each sample ID, each containing the three versions (unfiltered, filtered, and processed) of the expression data as well as the summary QC report and cell type report all named according to the ScPCA library ID.
+The `single_cell_metadata.tsv` file contains sample metadata for all samples included in the download.
+The `README.md` file provides information about the contents of each download file, additional contact and citation information, and terms of use for data downloaded from the ScPCA Portal.
+The files `bulk_quant.tsv` and `bulk_metadata.tsv` are only present for projects that also have bulk RNA-Seq data and contain, respectively, a gene by sample matrix of raw gene expression as quantified by `salmon`, and associated metadata for all samples with bulk RNA-Seq data.
+B. File download structure for an ScPCA Portal merged project download in `SCE` format.
+The download folder is named according to both the project ID and the date it was downloaded.
+Download folders contain a single merged object containing all samples in the given project as well as a summary report briefly detailing the contents of the merged object.
+All summary QC and cell type reports for each individual library are also provided in the `individual_reports` folder arranged by their sample ID.
+As in panel (A), additional files `single_cell_metadata.tsv`, `bulk_quant.tsv`, `bulk_metadata.tsv`, and `README.md` are also included.
+C. Overview of the merged workflow.
+Processed `SCE` objects associated with a given project are merged into a single object, including ADT counts from CITE-seq data if present, and a merged summary report is generated.
+Merged objects are available for download either in `SCE` or `AnnData` format.
+D. Example of UMAPs as shown in the merged summary report.
+A grid of UMAPs is shown for each library in the merged object, with cells in the library of interest shown in red and all other cells belonging to other libraries shown in gray.
+The UMAP is constructed from the merged object such that all libraries contribute an equal weight, but no batch correction was performed.
+The libraries pictured are a subset of libraries in the ScPCA project `SCPCP000003`.
+
+
+![**Figure 4. Cell type annotation `scpca-nf` subworkflow.**](https://raw.githubusercontent.com/AlexsLemonade/scpca-paper-figures/main/figures/compiled_figures/pngs/figure_4.png?sanitize=true){#fig:fig4 width="7in"}
+
+A. Expanded view of the cell type annotation subworkflow in `scpca-nf`, as introduced in Figure 2A.
+Cell type annotation is performed on the `Processed SCE Object`.
+A `celldex` [@doi:10.1038/s41590-018-0276-y] reference dataset with ontology labels is used as input for annotation with `SingleR` [@doi:10.1038/s41590-018-0276-y], and a list of marker genes compiled from `PanglaoDB` [@doi:10.1093/database/baz046] is used as input for annotation with `CellAssign` [@doi:10.1038/s41592-019-0529-1].
+Results from cell type annotation are then added to the `Processed SCE Object`, and a cell type summary report with information about reference sources, comparisons among cell type annotation methods, and diagnostic plots is created.
+Although not shown in this panel, cell type annotations are also included in the `Processed AnnData Object` created from the `Processed SCE Object` (Figure 2A).
+
+B. Example heatmap as shown in the cell type summary report comparing annotations with `SingleR` and `CellAssign`.
+Heatmap cells are colored by the Jacard similarity index.
+A value of 1 means that there is complete overlap between which cells are annotated with the two labels being compared, and a value of 0 means that there is no overlap between which cells are annotated with the two labels being compared.
+The heatmap shown is from library `SCPCL000498`.