Single-Cell Analysis in Veterinary Research

Practical Experience in Cross-Species Studies Using BD Rhapsody

Jing-Yuan Chen, DVM, PhD
jychen@niar.org.tw

National Center for Biomodels
National Institutes of Applied Research

August 21, 2026

A Veterinarian

What my granny thinks I do

What I thought I would do

What my friends think I do

The real me

Who Am I?

  • A veterinarian
  • PhD training in veterinary pathology
    • Animal infectious diseases
    • Swine vaccine: PCV2, PRRSV
    • Avian diseases: PaBV
  • A newcomer to single-cell analysis

Parrot Was the Starting Point

The Parrot Market

  • Import

  • Export

Data source: 財政部關務署 / 海關進出口統計

What Is PaBV?

Parrot bornavirus (PaBV)

  • Bornaviridae
  • Orthobornavirus
    • Mammals
    • Birds
    • Snakes
  • Genotypes
    • Alphapsittaciforme
      PaBV-1, 2, 3, 4, 7
    • Betapsittaciforme
      PaBV-5 and -6

Reference: Rubbenstroth, 2022, Avian Bornavirus Research - A Comprehensive Review

Anatomy of a Parrot

Parrot Bornavirus Infection

Reference DOI:
1. 10.3390/v14071513
2. 10.3390/v14102181

Summary of PaBV Studies

Sources: Leal de Araujo et al., 2017; Gartner et al., 2020/2021; Gray et al., 2010; Gancz et al., 2009; Mirhosseini et al., 2011; Escandon, 2015; Olbert et al., 2016; Runge et al., 2017; Rall et al., 2019; Murray et al., 2017.

What Was Still Missing?

  • Pathogen detection, pathology, and clinical records were available

  • Pending questions

    • Target cells?
    • The host responses?

  • What we needed next for

    • Vaccines
    • Theropies

Why We Returned to Parrots

Cell systems / mouse models

  • Viral replication
  • Feasibility testing
  • Early mechanistic clues

Still missing

  • Psittacine host specificity
  • Nervous-tissue disease
  • Disease progression

  • PaBV research is still restricted to the psittacine host context.

Parrot Housing

Our Parrot Trial

Groups: Control PBS, n = 2 · PaBV-4, n = 2 · PaBV-5, n = 2

  • Baseline — Screening
  • 0 dpi — IM inoculation
  • 14 dpi — Booster
  • 0-28 dpi — Observation
  • 28 dpi — Necropsy and tissue collection

Terminal workflow:
Histopathology / IHC → cerebrum and cerebellum dissociation → scRNA-seq

Moving Into Advanced Biomedical Approaches

What We Expected Form a Single-Cell Experiment

  • Add cell-level information to a short animal experiment
  • Find candidate host-response signals worth following
  • Keep useful clues for the next experiment

BD Rhapsody Workflow: From Tissue to Data

Taking scRNA-seq Beyond Standard Model Species

Sample prep

  • Dissociation and viability vary by tissue.

Capture / loading

  • Fragile samples limit usable cell input.

Library strategy

  • Pooling and tags shape what can be compared.

Sequencing

  • Depth and balance affect how much signal is usable.

Data analysis

  • References and gene symbols are incomplete.

Validation

  • Antibodies and pathology support interpretation.

Sample Preparation and Viability QC

BD checkpoint Parameter / guidance
Timing Prepare cells close to cartridge loading
Viability >50% recommended minimum
Wash steps Expect 25-50% cell loss
Cell size >20 µm may reduce bead loading efficiency
Cell concentration Dilute >1,000 cells/µL
to ~200-500 cells/µL
WTA cell load Intended total load: 1,000-20,000 cells
  • Not just protocol numbers
  • They shape which cells reach the cartridge

Sources: BD Doc ID 210964; BD Doc ID 23-22951; BD Doc ID 23-24117.
Picture: http://www.lsrc.u-toyama.ac.jp/mgrc/mgrc_inst_singlcell.htm

Pooling, Sample Tags, or Separate Libraries

BD offers direct and Flex sample-tagging routes.

Option Tag target Capacity
Human SMK
Cat. No. 633781		 Human universal Ab Up to 12-plex
Mouse Immune SMK
Cat. No. 633793		 Anti-mouse CD45 Up to 12-plex
Custom SMK
Cat. No. 626545		 Anti-mouse MHC-H2 Class I Up to 12-plex
Flex SMK A-D
Cat. No. 633849-633852		 Anti-PE + PE-primary Ab Up to 24-plex
  • Plexing capacity is the tagging limit
  • Practical pooling number depends on cell recovery, sequencing depth, and interpretability.

Sources: BD Single-Cell Multiplexing Kit product page; BD Doc ID 23-21340; BD Doc ID 23-24311.

The Missing Piece Was a Parrot-Compatible Antibody

  • Flex SMK can use PE-primary antibodies
  • But the antibody still has to label the target cells reliably
  • For parrot samples, finding a suitable antibody was difficult.

  • Separate libraries may be less efficient but easier to interpret in animal samples.

Third-Party Multiplexing Expands the Design Space

MULTI-seq concept

  • Lipid anchors hold DNA sample barcodes on the membrane
  • No species-specific antibody is required
  • Compatibility still needs user-side validation

Adapted from: Sigma-Aldrich MULTI-seq technical article;
McGinnis et al., Nature Methods 2019, doi:10.1038/s41592-019-0433-8.

WTA Library Preparation Workflow

WTA library workflow

  1. Captured cDNA on beads
  2. Random priming and extension
  3. RPE PCR
  4. Cleanup and quantification
  5. WTA index PCR
  6. Cleanup, QC, and sequencing

Protocol family

Route BD document
WTA Next 23-24991
WTA + AbSeq 23-24992
WTA + Sample Tag 23-24993
WTA + Sample Tag + AbSeq 23-24994

Sources: BD Doc ID 23-24117; BD Doc IDs 23-24991 to 23-24994.

BD Rhapsody Data Analysis Workflow

  • Seven Bridges handled the BD pipeline step
  • User decisions returned at filtering, annotation, and interpretation.

Sources: BD Rhapsody Sequence Analysis Pipeline User’s Guide, Doc ID 23-24580; Li et al., 2024, Current Protocols, doi:10.1002/cpz1.963.

The Reference Genome Was Another Gap

What BD needed

  • A BD-compatible archive: .tar.gz
  • Indexed genome: STAR index
  • Transcript annotation: GTF
  • Matching FASTA + GTF source
    • GENCODE for curated human / mouse annotation

Where I got stuck

  • No ready-made parrot reference
  • Genome FASTA existed, but GTF was not ready
  • GBFF could be converted, but still needed validation
  • Failed reference builds became part of the workflow
  • For rare species, reference preparation is not just a technical step
  • It shapes what can be interpreted later.

Source: BD Rhapsody Sequence Analysis Pipeline User Guide, Doc ID 23-24580.

From Failed Builds to a Reference Library


What this forced me to build

Reference group Examples
Standard models Human, mouse, zebrafish, rat
Veterinary species Chicken, quail, swine
Non-model species Budgerigar, dolphin, grampus, soybean
Project-specific PaBV-4, PaBV-5

From Sequencing to a First tSNE Map

  • A first structure emerged from the sequencing data

  • The map suggested separable cell populations

  • Biological meaning still depended on annotation

Cell Annotation

Clusters need evidence before they become cell labels

Approach Typical tools / resources What it helps with Main limitation
Canonical markers Literature, marker genes, feature plots Transparent biological reasoning Depends on marker conservation
Reference mapping SingleR, Azimuth, scmap, Seurat label transfer Fast transfer from curated atlases Only works as well as the reference
Gene-set enrichment Enrichr, GO, pathway / cell-type libraries Interpreting marker lists Supportive, not definitive
  • Annotation is an evidence-weighted biological decision.

Sources: BD Rhapsody Sequence Analysis Pipeline User Guide, Doc ID 23-24580; Enrichr gene set enrichment platform; Li et al., 2024, Current Protocols, doi:10.1002/cpz1.963.

How I Annotated Cockatiel Brain Cells

Canonical markers

Conservative annotation

  • No mature cockatiel brain atlas
  • Canonical marker sets
  • Broad cell classes first
  • Unresolved clusters stay unresolved
  • In a non-model species, a cautious label is stronger than an over-specific one.

From Genes Back to Pathology

The validation loop matters in non-model animals

  • Use scRNA-seq to nominate genes and cell populations
  • Translate candidate genes into possible protein markers
  • Check antibody availability and validate with IHC/pathology

My Practical Lessons From The Story

  • For veterinary samples, the value comes from keeping the animal model, sample quality, reference, and validation connected.

The Field Moved Forward While We Were Learning

Same dataset, richer context

At the time

  • Few animal scRNA-seq examples
  • Limited reference context
  • Manual marker validation

Field expands

  • More veterinary and livestock datasets
  • Emerging poultry and atlas resources
  • More spatial follow-up examples

Back to the data

  • Stronger annotation scaffold
  • Better marker prioritization
  • Clearer validation planning
  • The early experiment remains exploratory
  • Interpretation can now be more resource-informed and robust.

References to integrate: Stuart et al., 2019; Ramarapu et al., 2024; Lyons et al., 2024; Lu et al., 2024; Weiderman et al., 2025; Wang et al., 2025.

FarmGTEx and OmiGA: New Resources for Animal Omics

Animal omics resources are becoming easier to search, compare, and reuse

FarmGTEx Project, Nature Genetics, 2025.
DOI: 10.1038/s41588-025-02121-5

FarmGTEx Project, Nature Genetics, 2025. DOI: 10.1038/s41588-025-02121-5; OmiGA: https://omiga.bio/

What Can OmiGA Help Users Explore?

From animal omics data to searchable analysis context

FarmGTEx Project, Nature Genetics, 2025; OmiGA: https://omiga.bio/

Data Analysis Tools Became More Accessible

From steep scripts to guided interpretation

R / packages
Flexible and reproducible, but the learning curve is steep.

SeqGeq
More graphical, but still difficult for users without bioinformatics training.

Cellismo + AI copilots
Lower the entry barrier and help users return to QC, interpretation, and the research question.

AI helps with code, checking, and explanation;
the researcher still owns validation.

AI-assisted single-cell analysis examples: InstructCell, arXiv:2501.08187; SOAR benchmark, arXiv:2412.02915.

NTU ARC: Animal Models and Sample-Side Support

From animal model to interpretable sample

  • Veterinary and technical support
  • Animal rooms, surgery, and behavior workflows
  • Consortium of Integrative Biomedical Science Key Technology

NCB: Animal Models and Preclinical Infrastructure

Organized animal resources make advanced biomedical workflows more usable

National-level support for animal resources, preclinical studies, and quality systems.

Animal resources
RMRC
animal supply
customized models

Preclinical studies
disease models
technical service
efficacy testing

QC and training
health monitoring
veterinary diagnosis
technical education

From animal model resources to interpretable preclinical data.

Connecting Animal Models With Multiomic Profiling

  • animal and tumor models
  • flow-based immune profiling
  • IHC / pathology support
  • scRNA-seq for cellular resolution
  • spatial gene expression

Example resource
biomodels.ncb.org.tw

Source: NCB PDX model bank, https://biomodels.ncb.org.tw/pdx/pdx/about

Acknowledgements

DVM, NTU

  • Professor Hui-Wen (Winni) Chen
  • Meng-Chi (Angle) Wu
  • Zih-Syun (Angela) Fang
  • Hao-Fen (Tiffany) Chuang

Animal Resource Center, NTU

National Center for Biomodels

  • Dr. Yu-Chia Su
  • Veterinary Diagnostic Division

Contact Me

陳敬元 DVM, PhD / NCB, NIAR

Contact QR code

✉: jychen@niar.org.tw

Thank You!