R package for Kraken2, Kaiju and Bracken reports: read them into a tidy table, or start from an abundance matrix, a phyloseq object, or a table exported from Python, then plot composition, diversity and ordination.
# install.packages("remotes")
remotes::install_github("dsmutin/aRchiteutis")
library(aRchiteutis)archi_report() reads a directory of Kraken, Kaiju or QIIME 2 profiles (or a taxid abundance table), builds a phyloseq object with a taxonomy tree, and drops samples that fall below a read floor or whose rarefaction curve is still rising. Each requested plot is written with a one-line method caption. The HTML and the MultiQC section both end by stating that this is a preliminary report, not a final analysis.
path <- system.file("extdata", package = "aRchiteutis")
legend <- system.file("extdata", "legend.csv", package = "aRchiteutis")
archi_report(
path, legend, outdir = "report",
pattern = "_k2\\.txt$", trim_char = "_", target = "stage",
plots = c("composition", "donut", "alpha", "beta", "rarefaction")
)source is "auto", "kraken", "kaiju", "qza" or "abundance". plots may also include barplot, heattree, upset and difftree. beta_method is any name from archi_beta_methods().
A rendered PDF of that report on the bundled Kraken2 profiles is examples/kraken2-report.pdf.
Agent skills in skills/ only call package functions. Start with skills/aRchiteutis/SKILL.md.
| Step | Skills |
|---|---|
| Load a tidy table | aRchiteutis-load-kraken, aRchiteutis-load-abundance, aRchiteutis-load-phyloseq, aRchiteutis-load-python |
| Import to phyloseq | aRchiteutis-import-kraken, aRchiteutis-import-kaiju, aRchiteutis-import-qza, aRchiteutis-import-abundance, aRchiteutis-taxonomy-tree |
| Trim and reshape | aRchiteutis-transform |
| Plot | aRchiteutis-plot-composition, aRchiteutis-plot-diversity, aRchiteutis-plot-rarefaction, aRchiteutis-plot-difftree, aRchiteutis-plot-heattree, aRchiteutis-plot-ordination |
| Preliminary report | aRchiteutis-report |
The same files are mirrored under .cursor/skills/ for the editor.
Kraken2 reports (example files ship in the package):
path <- system.file("extdata", package = "aRchiteutis")
legend <- system.file("extdata", "legend.csv", package = "aRchiteutis")
df <- get_counts(path, pattern = "decont_b", legend = legend, trim_char = "_")Other inputs, same plotting functions:
df <- from_abundance(count_matrix, clade = "G") # taxa in rows, samples in columns
df <- from_phyloseq(physeq, taxa_rank = "Genus")
df <- python2r("abundance.csv", clade = "G") # CSV from inst/python/python2r.pyThe table is long: taxa, clade, sample, N, amount, amount_cl, plus legend columns. df_untidy() turns it into a taxa-by-sample matrix. Trim with df_taxa_trim() or df_get_top_taxa() before plotting. Argument details and a second copy of every figure are in vignettes.md.
Figures below are rendered from the bundled honey-bee brood reports.
df2donut() — mean composition.
df2composition() — stacked bars per sample. Samples are ordered by the first principal component (order_samples = "fpc"). hclust, abundance, alpha and none are the other orders.
df2barplot() — amount per taxon. style = "raincloud" draws a ggviolinbox raincloud.
df2composition_tree() — fan cladogram from ggtree (Yu et al. 2017) with a ggtreeExtra boxplot of relative abundance (Xu et al. 2021). Tips and boxes are coloured by phylum when a rank table is available.
df2alpha_summary() — alpha-diversity indices. split_by is a legend column (here, brood stage).
df2alpha() — Shannon and Simpson. style = "raincloud" uses the same raincloud geometry.
df2rarefaction() — observed richness, Shannon and Simpson against sequencing depth. One thin line per sample, a smooth by group.
df2beta() — Bray–Curtis distances between samples. Point colour is brood stage. method accepts any name from archi_beta_methods(), including "aitchison".
df2beta_pcoa() — PCoA of that distance, with group ellipses.
df2heattree() — metacoder heat tree. Node size is the number of taxa, colour is mean relative abundance. Empty ranks are dropped and their children reattached, so the taxonomy is one tree. The function stops when \pkg{metacoder} is not installed.
df2difftree() — log2 fold change between two groups on a ggtree layout. The default fruit is a heatmap of relative abundance; tip colour is the fold change. fruit = "bar" draws the fold change as a column. With more than two groups, pass contrast. engine = "metacoder" draws metacoder::heat_tree and stops when that package is not installed. engine = "microbiota" runs MicrobiotaProcess mp_diff_analysis and stops when that package is not installed.
engine = "microbiota" draws the MicrobiotaProcess biomarker tree (radial ggtree, phylum highlight, abundance stars, LDA and FDR) when that package is installed.
df2upset() — taxon presence as a ComplexUpset plot. The function stops when \pkg{ComplexUpset} is not installed.
df2cluster() — dendrogram of taxa or of samples. Leaf colour is the cluster.
df2clust2d() — mean abundance in two groups on a log10 scale (pupa vs larvae).
df2heatmap() — taxa by sample.
df2corrplot() — correlations between taxa.
df2chord() — circular co-occurrence graph.
df2tsne() — t-SNE of taxa.
df2volcano() — ANCOM-BC2 log2 fold change against adjusted q. The first legend_detect pattern is the reference group. The function stops when \pkg{ANCOMBC} is not installed.
df2pca_sample() — PCA of samples. df2pca_sp() — PCA of taxa.
citation("aRchiteutis")The record lives in inst/CITATION: Smutin, Taldaev, Lebedev & Adonin, International Journal of Molecular Sciences 25(2):741, 2024. https://doi.org/10.3390/ijms25020741