Mixed-species scRNA-seq analysis guide: principles interpretation and common issues

Author: SeekGene

Time: 11 min

Words: 2.0k words

Updated: 2026-01-22

Reads: 0 times

FAQ

NOTE

Important Prerequisites

SeekSoul Tools does not have species-specific algorithms for mixed species samples. Mixed species samples should be converted to 10x whitelist format and analyzed using Cell Ranger
Multi-omics ATAC does not support mixed species analysis: 10x software cellranger arc and cellranger ATAC do not support mixed species. This guide applies only to standard RNA libraries
Barcode conversion tool: Refer to the Barcode-Converter utility in the bioinformatics external resources on DingTalk

Building Mixed Species Reference Genome

Data Preparation

NOTE

File Format Requirements

Recommended: Use GTF and FASTA files from the Ensembl database (includes optional tags for subsequent filtering)
If not available in Ensembl, GTF and FASTA files from other sources can be used
GTF format is required; GFF format is not supported
For GTF/GFF format specifications, refer to: Ensembl GFF/GTF Format Documentation

Filtering Non-coding Genes (Recommended)

GTF files may contain entries related to non-polyA transcripts that overlap with protein-coding genes. Due to these overlapping annotations, reads may be marked as mapping to multiple genes (multi-mapped) and will not be used for subsequent gene quantification.

It is recommended to use the cellranger mkgtf command to filter non-coding genes:

shell

cellranger mkgtf input.gtf output.gtf --attribute=gene_biotype:protein_coding

Building Mixed Species Index

Use the cellranger mkref command to build a mixed species reference genome. You need to specify --genome, --fasta, and --genes parameters for each species:

shell

cellranger mkref \
  --genome=GRCh38 \
  --fasta=/human/refdata-gex-GRCh38-2020-A/fasta/genome.fa \
  --genes=/human/refdata-gex-GRCh38-2020-A/genes/genes.gtf \
  --genome=Macaca_fascicularis \
  --fasta=/Macaca_fascicularis_GCF_037993035.1/fasta/genome.fa \
  --genes=/Macaca_fascicularis_GCF_037993035.1/genes/genes.gtf \
  --maxjobs 50 \
  --localmem 60 \
  --localcores 12

Verifying Index Construction

After construction, check the reference.json file to ensure it contains 2 genome entries. Cell Ranger will only recognize it as a mixed species reference if this condition is met:

json

{
    "fasta_hash": "6a2ba19a695f279ec863f521fb1d834d01905c19",
    "genomes": [
        "GRCh38",
        "Macaca_fascicularis"
    ],
    "gtf_hash.gz": "24cb2b8556e5c3740b0816f965c7912f679b17f5",
    "input_fasta_files": [
        "genome.fa",
        "genome.fa"
    ],
    "input_gtf_files": [
        "genes.gtf",
        "genes.gtf"
    ],
    "mem_gb": 16,
    "mkref_version": "8.0.0",
    "threads": 2,
    "version": null
}

NOTE

Handling Exogenous Inserted Genes

If you have exogenous inserted genes, you can first add these genes to the genome and annotation files of any species, then build the mixed species index.

Analysis Workflow and Principles

Core Principle Overview

Cell Ranger classifies species based on UMI counts mapped to different species for each barcode:

Initial Classification: Barcodes are initially classified to the species with higher UMI counts
Multiplet Identification: If a barcode has UMI counts exceeding the 10th percentile of each species' distribution, it is identified as a multiplet (multiple cells mixed together)

NOTE

Algorithm Applicability

This identification algorithm is more accurate when the mixed species cell ratio is close to 1:1 (e.g., 50:50). Accuracy decreases as the ratio deviates further from this balance.

Detailed Algorithm Steps

Step 1: Collect and Analyze UMI Counts

For each cell barcode, Cell Ranger counts:

Species A UMI Count: Number of UMIs mapped to Species A reference genome
Species B UMI Count: Number of UMIs mapped to Species B reference genome

Based on this, each barcode can be classified into one of three categories:

Species A Dominant (A > B): More UMIs mapped to Species A than Species B
Species B Dominant (B > A): More UMIs mapped to Species B than Species A
Comparable or Mixed (A ≈ B): Possibly a dual-species mixture

Step 2: Setting Classification Threshold

To distinguish "true single cells" from "possible multiple cells", Cell Ranger uses the 10th percentile as the classification threshold:

Setting Species A Cell Threshold:
- From all barcodes where "Species A UMI > Species B UMI"
- Find the 10th percentile of Species A UMI counts for these barcodes, denoted as A_threshold
- Any barcode with Species A UMI count > A_threshold is initially considered a Species A singlet
Setting Species B Cell Threshold:
- From all barcodes where "Species B UMI > Species A UMI"
- Find the 10th percentile of Species B UMI counts for these barcodes, denoted as B_threshold
- Any barcode with Species B UMI count > B_threshold is initially considered a Species B singlet

Step 3: Identifying Multiplets

A barcode is identified as a multiplet if it simultaneously satisfies both conditions:

Species A UMI count > A_threshold AND
Species B UMI count > B_threshold

Algorithm Workflow Summary

typescript

For each barcode:
    Count speciesA_UMI_count and speciesB_UMI_count
    
    If speciesA_UMI > speciesB_UMI:
        Classify as "Species A Dominant"
        → Use all "Species A Dominant" barcodes, take 10th percentile of speciesA_UMI → A_threshold
    
    If speciesB_UMI > speciesA_UMI:
        Classify as "Species B Dominant"
        → Use all "Species B Dominant" barcodes, take 10th percentile of speciesB_UMI → B_threshold

Judgment Logic:
    If speciesA_UMI > A_threshold AND speciesB_UMI > B_threshold:
        → Identified as multiplet (dual-species mixture)
    If only speciesA_UMI is high → Possibly Species A singlet
    If only speciesB_UMI is high → Possibly Species B singlet
    (Note: Cannot distinguish dual cells of the same species)

Interpreting Results

Summary Report

The Summary report provides the following key information:

Mapping rates for each species: Reflects data quality
Number of cells identified by software: Number of cells detected for each species
Median genes: Median gene expression per cell

Gene Expression Report

Key Metrics Explanation

Metric	Definition	Interpretation
GEMs with >0 Cell	Number of barcodes with at least one cell after identification	Represents the number of valid cells detected
GEMs with >1 Cell	Number of barcodes with at least 2 cells after identification	Represents the number of multiplets. A high proportion indicates that some barcodes may contain RNA from multiple cells
Fraction GEMs with >1 Cell	Estimated average proportion of barcodes associated with multiple cells	Calculated through bootstrap sampling and adjusted for the ratio of two cell types. Reflects the proportion of barcodes inferred to be associated with multiple cells among all barcodes
Cell UMI Counts Plot	Each point represents a barcode, with axes measuring total UMI counts mapped to each species	Point colors indicate the inferred number of cells associated with each barcode, providing a visual representation of which barcodes may be multiplets

Output Files

Standard Output Files

Consistent with standard RNA analysis results, including:

filtered_feature_bc_matrix/: Filtered expression matrix
raw_feature_bc_matrix/: Raw expression matrix
analysis/: Analysis results directory
metrics_summary.csv: Quality control metrics summary

Special File: `gem_classification.csv`

This file records the UMI counts for each species detected for each barcode and the final species classification determined by the software:

Column Descriptions:
- barcode: Cell barcode
- [Species A Name]: UMI count mapped to Species A
- [Species B Name]: UMI count mapped to Species B
- call: Species classification determined by software (Species A name, Species B name, or Multiplet)

Example:

csv

barcode,GRCh38,Rattus,call
AAACCCAAGAAACACT-1,44666,22,GRCh38
AAACCCAAGAAACCAT-1,5856,8,GRCh38
AAACCCAAGAAACCCA-1,35589,10,GRCh38
AAACCCAAGAAACCCG-1,37726,13,GRCh38
AAACCCAAGCGATCTC-1,296,14235,Rattus
AAACCCAAGCGCACTT-1,271,15742,Rattus
AAACCCAAGGCGGACT-1,276,12822,Rattus
AAACCCAAGGCTCACC-1,246,13719,Rattus
AAACCCACAATCCCTC-1,2978,8727,Rattus
AAACGAAAGTGAGTGC-1,4600,839,Multiplet
AAACGAAAGTGATGGC-1,3827,983,Multiplet
AAACGAAAGTGCGATA-1,5605,907,Multiplet
AAACGAACAAACCCGC-1,4674,940,Multiplet

Downstream Analysis: Species Cell Identification

There are two analytical approaches for species cell identification and separation:

Method 1: Using Cell Ranger Classification Results

Directly use the species classification determined by Cell Ranger software in gem_classification.csv:

# Read Cell Ranger classification results
cellranger <- read_csv("gem_classification.csv") %>%
  select(barcode, call) %>%
  rename("cellranger" = "call")

cellranger$barcode <- as.character(cellranger$barcode)

# Merge into Seurat object
ob@meta.data <- ob@meta.data %>%
  tibble::rownames_to_column(var = "barcode") %>%
  left_join(cellranger, by = "barcode") %>%
  tibble::column_to_rownames(var = "barcode")

Method 2: Judging Based on Gene Expression Percentage

Determine species based on the percentage of genes from each species expressed in each cluster:

# Calculate gene expression percentage for each species
ob[['percent.Rat']] <- PercentageFeatureSet(object = ob, pattern = '^Rattus')
ob[['percent.Human']] <- PercentageFeatureSet(object = ob, pattern = '^GRCh38')

# Visualization
p <- DimPlot(ob, label = T, repel = T, label.box = T)
p1 <- VlnPlot(ob, features = c("percent.Rat", "percent.Human"), pt.size = 0)
options(repr.plot.height = 6, repr.plot.width = 16)
p + p1

NOTE

Cloud Platform Tool Support

The cloud platform's "My Tools" includes a SpeciesSplit module that can assist with species separation analysis.

Common Issues and Troubleshooting

Q1: How to determine if the analysis results contain cells from the corresponding species

Methods for Determination:

Check mapping rates in the quality control report:
- Examine the mapping rate (Mapping Rate) for each species
- If a species' mapping rate is very low (e.g., < 5%), it suggests that cells from that species may not have been captured
Check UMI and gene counts:
- If a species' mapping rate is low, the detected UMI and gene counts will be very low
- The final identified cell count for that species is unreliable, indicating that cells from the corresponding species may not have been captured in the sample
Examine gem_classification.csv:
- Check the UMI count distribution for each species
- If a species' UMI counts are generally very low, it suggests that cells from that species may not exist

Troubleshooting Steps:

# Read classification results and check
classification <- read_csv("gem_classification.csv")

# Count cells for each species
table(classification$call)

# Check UMI count distribution for each species
summary(classification$GRCh38)  # Replace with actual species name
summary(classification$Rattus)  # Replace with actual species name

Q2: How to understand Multiplet identification in the sample

Key Points:

Algorithm Limitations:
- Cell Ranger provides a fixed algorithm based on the 10th percentile threshold
- Accuracy is higher when cell ratios are close to 1:1, and decreases as the ratio deviates further
Comprehensive Judgment Methods:
- Combine with overall clustering results of the sample
- Check gene expression patterns of these cells
- Calculate the expression percentage of corresponding species genes
- Examine overall gene and UMI counts for these cells
- Can be combined with downstream doublet prediction software (such as DoubletFinder, scDblFinder, etc.) for verification

Verification Code Example:

# Extract multiplet cells
multiplet_cells <- rownames(ob@meta.data)[ob@meta.data$cellranger == "Multiplet"]

# Check gene expression characteristics of multiplet cells
multiplet_data <- ob@meta.data[multiplet_cells, ]
summary(multiplet_data$nFeature_RNA)
summary(multiplet_data$nCount_RNA)

# Check species gene expression percentage
if ("percent.Human" %in% colnames(ob@meta.data)) {
  summary(multiplet_data$percent.Human)
  summary(multiplet_data$percent.Rat)
}

Q3: Can sample data analyzed using a mixed species reference genome be integrated with samples analyzed using a single species reference genome

Answer: Yes, but preprocessing is required.

Reason:

When using a mixed species reference genome, both species' genomes are tagged. For example, species prefixes are added to all gene names:

Human genes: GRCh38_CD79A
Mouse genes: mm10_Actb

Integration Steps:

Split Expression Matrix:
- Split the expression matrix based on species prefixes
- Separate mixed species data into two single-species expression matrices
Rename Genes:
- Remove species prefixes to restore original gene names
- For example: GRCh38_CD79A → CD79A
Integrate Separately:
- Integrate human cells with single-species human data
- Integrate mouse cells with single-species mouse data

Example Code:

# Read mixed species data
mixed_seurat <- Read10X("mixed_species_output/")

# Split human and mouse data
human_genes <- grep("^GRCh38_", rownames(mixed_seurat), value = TRUE)
mouse_genes <- grep("^mm10_", rownames(mixed_seurat), value = TRUE)

# Extract human cells (based on gem_classification.csv)
human_cells <- classification$barcode[classification$call == "GRCh38"]
human_matrix <- mixed_seurat[human_genes, human_cells]

# Rename genes (remove prefix)
rownames(human_matrix) <- gsub("^GRCh38_", "", rownames(human_matrix))

# Create Seurat object and integrate with single-species data
human_seurat <- CreateSeuratObject(human_matrix)
# ... subsequent integration analysis

Last updated: November 2025

Mixed-species scRNA-seq analysis guide: principles interpretation and common issues ​

Building Mixed Species Reference Genome ​

Data Preparation ​

Filtering Non-coding Genes (Recommended) ​

Building Mixed Species Index ​

Verifying Index Construction ​

Analysis Workflow and Principles ​

Core Principle Overview ​

Detailed Algorithm Steps ​

Step 1: Collect and Analyze UMI Counts ​

Step 2: Setting Classification Threshold ​

Step 3: Identifying Multiplets ​

Algorithm Workflow Summary ​

Interpreting Results ​

Summary Report ​

Gene Expression Report ​

Key Metrics Explanation ​

Output Files ​

Standard Output Files ​

Special File: gem_classification.csv ​

Downstream Analysis: Species Cell Identification ​

Method 1: Using Cell Ranger Classification Results ​

Method 2: Judging Based on Gene Expression Percentage ​

Common Issues and Troubleshooting ​

Q1: How to determine if the analysis results contain cells from the corresponding species ​

Q2: How to understand Multiplet identification in the sample ​

Q3: Can sample data analyzed using a mixed species reference genome be integrated with samples analyzed using a single species reference genome ​

Mixed-species scRNA-seq analysis guide: principles interpretation and common issues

Building Mixed Species Reference Genome

Data Preparation

Filtering Non-coding Genes (Recommended)

Building Mixed Species Index

Verifying Index Construction

Analysis Workflow and Principles

Core Principle Overview

Detailed Algorithm Steps

Step 1: Collect and Analyze UMI Counts

Step 2: Setting Classification Threshold

Step 3: Identifying Multiplets

Algorithm Workflow Summary

Interpreting Results

Summary Report

Gene Expression Report

Key Metrics Explanation

Output Files

Standard Output Files

Special File: `gem_classification.csv`

Downstream Analysis: Species Cell Identification

Method 1: Using Cell Ranger Classification Results

Method 2: Judging Based on Gene Expression Percentage

Common Issues and Troubleshooting

Q1: How to determine if the analysis results contain cells from the corresponding species

Q2: How to understand Multiplet identification in the sample

Q3: Can sample data analyzed using a mixed species reference genome be integrated with samples analyzed using a single species reference genome