You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Dependencies
In DESCRIPTION,
build: require R 4.1.0 or newer, ggplot2 4.0.0 or newer, ggforce 0.5.0 or newer, ggrepel 0.9.6 or newer, and shiny 1.7.4 or newer, which pulls in fontawesome 0.4.0 and its Font Awesome 6 icon names.
build: require pandoc 2.11 or newer, which has citeproc built in, so pandoc-citeproc is no longer needed.
build: require knitr 1.39 or newer, which the report needs for include_graphics(rel_path = FALSE).
build: require scales 1.4.0 or newer, which ggplot2 4.0.0 already needs, for the log-10 axes of the "PFNF" and "HF" plots.
Chores
In inst/app/,
refactor: replace the superseded shiny::callModule() with shiny::moduleServer().
refactor: use the Font Awesome 6 icon names file-arrow-up and circle-info.
In R/,
refactor: drop stringsAsFactors = FALSE from data.frame() and as.data.frame() calls, where it is the default since R 4.0.0.
refactor: hide boxplot outliers with outliers = FALSE instead of outlier.shape = NA.
Documentation
In vignettes/, README.Rmd and the render() report template,
docs: write chunk options as #| YAML comments, which needs knitr 1.35 or newer.
docs: add alternative text to the images.
docs: install the development version with pak::pak() instead of remotes::install_github().
docs: fix typos and grammar, and spell GitHub, NanoString, R Markdown and Shiny consistently.
In vignettes/,
docs: keep only the GSE70970 samples measured with the NS_H_miR_1.4 CodeSet, since its two CodeSets come from different nSolver versions and load_rcc() refuses files that mix versions.
docs: attach data.table in the analysis vignette, which uses dcast() and as.data.table().
docs: link to GEO over https in the bibliography, since the http address now redirects.
In pkgdown/,
docs: restyle the website for pkgdown 2.2 with the NACHO logo colours, a light and dark mode switch, and colour contrast that meets WCAG AA.
docs: group the reference index by task.
In R/normalise.R, R/load_rcc.R and vignettes/NACHO.Rmd,
docs: remove the raw_counts and normalised_counts slots, which never existed, and say that nacho holds one row per sample and probe.
In R/GSE74821.R,
docs: say that GSE74821 holds 48 samples, not 20.
Fixes
In inst/CITATION,
fix: list the authors as person() objects, so the citation shows their full initials and spells Leen M. 't Hart correctly.
fix: cite the version of record, Bioinformatics volume 36, issue 3, pages 970 to 971, published in February 2020.
In R/autoplot.R,
fix: stop boxplots inheriting the colour aesthetic, which made ggplot2 warn that it dropped colour for the control probe plots.
fix: draw the outlier bands of the "PFNF" and "HF" plots to the panel edges without log-10 warnings about infinite values.
fix: keep each sample's identifier in the "BD", "FoV", "PCL", "LoD", "PN", "Positive", "Negative" and "Housekeeping" plots, which drew every point at one position and dropped some of them.
In R/render.R,
fix: include the logo by its absolute path, so pandoc finds it when the temporary directory sits behind a symbolic link.
fix: pass the report options as R Markdown parameters, so outliers_labels = "CartridgeID" works and column names with quotes no longer break the report.
In inst/app/app.R,
fix: stop writing an all.rdata debug file to the working directory when uploading RCC files.
fix: load uploaded RCC files, which failed because suppressMessages() received x instead of expr.
fix: make the sample sheet optional again when uploading RCC files, instead of failing on a missing ssheet_dt.
In DESCRIPTION,
fix: suggest markdown, which the app needs to show its help pages, and make visualise() ask for it when it is not installed. The deploy() help page notes that the server needs it too.
In R/geometric_housekeeping.R,
fix: replace background-corrected housekeeping counts below 1 with 1, so values between 0 and 1 no longer inflate House_factor. (#53)
In R/qc_pca.R,
fix: compute the PCA with samples as observations and store their scores. PC01 to PC10 and the variance explained change, and the PCA plots now show the main sources of variation between samples.
In R/normalise_counts.R,
fix: apply the 0.1 floor after rounding, so normalised counts at or below background are 0.1 instead of 0. With housekeeping_predict = TRUE, the floor can change which housekeeping genes are picked, which in turn changes House_factor, the normalised counts and the outlier flags.
In R/qc_pca.R and R/normalise_counts.R,
fix: objects created with an earlier version keep the old values, so run load_rcc() again to refresh them.
In R/normalise.R,
fix: use the n_comp passed to normalise(), which was ignored.
fix: recompute is_outlier when only the thresholds change.
In R/qc_features.R, R/qc_limit_detection.R and R/check_outliers.R,
fix: report PCL and LoD as NA when a panel has no POS_E probe or when the negative controls do not vary, and do not flag a sample on a metric that could not be measured.
In R/load_rcc.R,
fix: detect PlexSet files from their content, and add plexset_idS1 to S8 when the sample sheet lists each file once.
fix: stop converting the caller's sample sheet to a data.table.
fix: stop with a clear message when RCC files mix PlexSet and single-sample files.
fix: stop when a single-sample sample sheet lists the same RCC file twice.
In R/print.R,
fix: return the object invisibly from print().
In data/,
fix: rebuild GSE74821 with the corrected PCA and normalised counts.
In inst/app/,
fix: set the full binding density range when switching between the MAX/FLEX and SPRINT presets.
fix: accept zip archives sent as application/zip, and .RCC, .rcc, .RCC.gz and .csv files in any case.
fix: stop warning about row names when unpacking a zip archive.
fix: show a notification when the sample sheet is discarded, instead of a warning in the R console.
fix: correct the typos in the card and outlier labels, and open the app on the QC metrics tab.