SpeciesSplit Mixed-Species Matrix Partitioning

Author: SeekGene

Time: 4 min

Words: 764 words

Updated: 2026-10-13

Reads: 0 times

SpeciesSplit performs species-level partitioning on single-cell expression matrices from multiplexed multi-species experiments, separating the pooled matrix into species-specific sub-matrices for downstream analysis of each species independently.

Applicable Scenarios

When cells from multiple species are pooled and loaded into the same GEM, alignment is performed against a combined reference genome. After standard analysis, gene identifiers in the resulting expression matrix carry species-specific prefixes (e.g., GRCh38-GAPDH, Sus_scrofa-ALB). SpeciesSplit leverages this feature to automatically detect species prefixes, assign each cell to its species of origin, and partition the data into independent Seurat objects.

Partitioning Principle

Cell Species Assignment

Cell species identity is determined based on the user-selected method:

Method	Assignment Logic	Required Parameters
seekone	Compute the UMI proportion of species-prefixed genes per cell (denominator includes only UMIs from species-prefixed genes); assign the cell to a species if its proportion > Threshold; otherwise label as Multiplet	Threshold
cellranger	Read the call column from the CellRanger gem_classification.csv file to directly assign species identity	gemClassification file path

Handling of Unprefixed Genes

Viral genes (e.g., HBx, SV40) or engineered genes (e.g., CAR, fluorescent proteins) incorporated into the combined reference genome may lack a species prefix. During partitioning, such genes cannot be assigned to any single species. The module treats unprefixed genes as shared genes, retaining them in every species-specific sub-matrix so that their expression can be examined across all species downstream.

To restrict an unprefixed gene to a specific species only, it must be placed under that species namespace during reference genome construction (e.g., GRCh38_EGFP will be retained exclusively in the human sub-matrix and will not appear in other species' sub-matrices).

Cloud Platform Parameters

Parameter	Description
Analysis Pipeline	SpeciesSplit
Task Name	Task name; must not duplicate any existing task name
Sample Info	Enter sample names; analysis batch and file OSS paths are auto-populated. Verify the correct analysis batch
Platform	Species assignment method, either seekone or cellranger
Threshold	UMI proportion threshold, range 0–1, default 0.95. Only applicable to the seekone method
Description	Task description, optional

Notes

WARNING

Gene identifiers must contain a species prefix (e.g., GRCh38-GAPDH, mm10-Gapdh); otherwise the module cannot resolve species identity, and all cells will be labeled as unknown with partitioning skipped.

• Only transcriptomic data types are supported; multi-omics data types are not yet supported.
• A single sample name may correspond to multiple analysis batches. The last batch is populated by default; verify and select the intended batch.
• Genomes field accuracy: It is recommended to use cellranger mkref --genome to specify a separate genome name for each species during reference genome construction.
• Unprefixed exogenous genes: Viral genes (HBx, SV40) or engineered genes (CAR, fluorescent proteins) without a species prefix are retained as shared genes in every species-specific sub-matrix after partitioning; if a species prefix is present (e.g., GRCh38_EGFP), the gene appears only in the corresponding species' sub-matrix.
• Species prefix limit: The module supports a maximum of 10 species prefixes; exceeding this limit will trigger an error and terminate execution (typically indicating incorrect gene naming conventions).

Supported Species Prefixes

The built-in species_mapping.json mapping file covers the following species:

Species	Prefix Examples	English Name	Latin Name
Human	GRCh38, Homo_sapiens, human	human	Homo sapiens
Mouse	GRCm39, Mus_musculus, mm10, mouse	mouse	Mus musculus
Pig	Sus_scrofa, Sus, pig	pig	Sus scrofa
Rhesus macaque	Mmul_10, Mmul, Macaca_mulatta, Rhesus_monkey	Rhesus monkey	Macaca mulatta
Fruit fly	Drosophila_melanogaster, fruit_fly	fruit fly	Drosophila melanogaster
Zebrafish	Danio_rerio, zebrafish	zebrafish	Danio rerio
Chicken	Gallus_gallus, chicken	chicken	Gallus gallus
Cattle	Bos_taurus, Bos, cattle	cattle	Bos taurus
Dog	Canis_lupus_familiaris, Canis_familiaris, dog	dog	Canis lupus familiaris
African clawed frog	Xenopus_laevis, african_clawed_frog	african clawed frog	Xenopus laevis
Nematode	Caenorhabditis_elegans, nematode	nematode	Caenorhabditis elegans
Budding yeast	Saccharomyces_cerevisiae, budding_yeast	budding yeast	Saccharomyces cerevisiae
Goldfish	gold_fish, goldfish	gold fish	Carassius auratus
Red crucian carp	CAred, Red_crucian_carp	Red crucian carp	Carassius auratus red var
Rabbit	Rattus_add_EGFP, Rattus, rabbit	rabbit	Oryctolagus cuniculus

If the reference genome prefix used in your experiment is not present in the mapping file, the module will automatically fall back to regex-based prefix detection.

Comparison of Assignment Methods

Aspect	seekone	cellranger
Assignment basis	UMI proportion of species-prefixed genes	CellRanger classification output file
Required parameter	Threshold (proportion threshold)	gemClassification file
Multiplet detection	Cells below threshold are automatically labeled	Dependent on CellRanger output
Compatible platform	SeekSoul Tools / SeekArc Tools	10x Chromium

Both methods share identical species detection and partitioning logic; they differ only in how cell species identity is determined.