Advanced layer-aware workflow

Songqi Duan

duan@songqi.org

This article demonstrates three Shennong behaviors that are important in modern multi-layer Seurat workflows:

• automatic species inference when species is not provided
• running analysis from a non-default count layer
• storing DE metadata in object@misc$de_results

library(Shennong)
library(dplyr)
library(knitr)
library(Seurat)

Automatic species inference

The count matrix is initialized without an explicit species argument, but Shennong still infers the species from feature names using hom_genes and common mitochondrial patterns.

knitr::kable(
  tibble::tibble(
    object = c("raw_counts", "pbmc_inferred"),
    species = c(inferred_species, sn_get_species(pbmc_inferred))
  )
)

object	species
raw_counts	human
pbmc_inferred	human

Layer-aware clustering

The object below is clustered from the decontaminated_counts layer rather than the default counts layer.

sn_plot_dim(
  pbmc_layered,
  reduction = "umap",
  group_by = "seurat_clusters",
  label = TRUE,
  show_legend = FALSE,
  title = "pbmc1k clustered from the decontaminated_counts layer"
)

Stored differential expression metadata

sn_find_de(..., return_object = TRUE) stores both the marker table and the analysis metadata required for later reuse in plotting and enrichment.

knitr::kable(de_summary)

field	value
schema_version	1.0.0
package_version	0.1.2
analysis	markers
group_by	seurat_clusters
rank_col	avg_log2FC
p_col	p_val_adj
n_genes	15690

Reusing stored markers

sn_plot_dot(
  pbmc_layered,
  features = "top_markers",
  de_name = "cluster_markers",
  n = 3
)