Skip to content

CapyBio FAQ

About CapyBio and the technology

What is CapyBio?

CapyBio is a decision layer for scientists and teams working with humanized models. We help scientists understand, improve, and ultimately build better human cells.

We think of human cells as infrastructure for modern biotech. They're increasingly the foundation for drug discovery, disease modeling, cell therapy, and many other applications. Our job is to make sure that infrastructure actually reflects the biology it's supposed to represent.

We run two computational platforms, Capybara™ and CellOracle™, on your single-cell RNA-seq data. Capybara shows how closely your cells match their primary human counterparts, and CellOracle shows what to change to improve them. You get an integrated report and a live consultation to walk through the results. If you need help upstream, we also offer wet lab services such as scRNA-seq and cell culture.

Who uses CapyBio, and what problems does it solve?

Anyone who works with humanized models: lab-made human cells, organoids, and tissues that stand in for human biology.

Some are developing those models. That includes researchers building iPSC-derived, reprogrammed, or engineered cells, organoids, and tissues, cell therapy developers, and teams creating new differentiation approaches.

Others use them for an application. That includes drug development teams doing therapeutic discovery and toxicity or efficacy testing, and companies making media, reagents, or supplements for better cell maintenance and differentiation.

Either way, CapyBio helps you answer one question: is this model good enough? How closely does it match your target, how mature is it, how pure is it, and what should you change to improve it?

What technology does CapyBio use to analyze cell identity?

We call it the “GPS for cell engineering”: the platform shows where your cells are, where they need to go, and how to get there. It combines two complementary technologies.

Capybara™ measures where your cells are. It scores each engineered cell against primary human reference data using quadratic programming, which gives a continuous, transcriptome-wide identity score instead of forcing every cell into a single discrete label. CellOracle™ shows how to get where you want to go. It uses gene regulatory network (GRN) modeling to rank which transcription factor perturbations are most likely to move cells toward a target fate.

Both methods are peer-reviewed (Kong et al., Cell Stem Cell 2022 (opens in new tab); Kamimoto et al., Nature 2023 (opens in new tab)) and were developed by CapyBio co-founders with backgrounds in single-cell biology and GRN modeling.

How is CapyBio's analysis different from standard single-cell RNA-seq analysis or cell type annotation tools?

The key difference is how Capybara handles ambiguous cells. Most cell type classifiers force every cell into the nearest discrete label, so cells in transition get misclassified or dumped into “unknown.” Capybara does three things differently:

  • It measures identity as a continuous, quantitative score rather than a forced discrete call.
  • It reports mixed-identity (“hybrid”) and unclassified cells explicitly instead of hiding them inside the nearest label, which captures partial reprogramming and transitional states that conventional annotation misses.
  • It benchmarks against curated primary human reference states, giving an objective measure of how faithfully engineered cells recapitulate the target cell type.

Can I get a quantitative score for how well my clusters match a reference cell type, instead of annotating with marker genes?

Yes, this is what Capybara is built for. In a typical single-cell workflow, cells are clustered and each cluster is labeled based on a few marker genes, which leaves open how good the match really is. Capybara assigns every cell a continuous identity score against the reference and explicitly reports hybrid and unclassified cells, so you see not just a label but how strongly each cell matches it.

If you already have scRNA-seq data, we start from your raw count matrix. If not, our sequencing add-on can generate it.

How does CapyBio use the reference sample, and how do I choose the best one for my project?

Capybara scores your engineered cells against a reference to calculate identity, purity, and off-target composition. Every package includes working with the CapyBio team to select the most appropriate reference for your project, so you don't have to choose alone. We have access to a wide variety of public datasets, including the Human Cell Atlas, CellxGene, and Tabula Sapiens, and build references around what you need.

We usually recommend a reference that spans the differentiation timeline, for example iPSC, primary germ layer or primitive progenitor, committed progenitor, and mature cell type. The more comprehensive the reference, the better Capybara is at classifying your cells. The 1-hour kickoff meeting included with every package is where we confirm your project goals and review your sample and reference details.

Primary cells vary between donors and change in culture. What are my cells actually being compared to?

It's a fair question. Primary human cells, the usual gold standard, vary from donor to donor and can lose their characteristic phenotype quickly once cultured. Capybara benchmarks against curated primary human reference data, not a single donor's cultured cells, so your results are anchored to a consistent, objective standard.

References are built around your needs and represent the ideal condition you're trying to replicate. That could be a fully functional hepatocyte with specific CYP450 activity, or a specific neuronal type from patients with Alzheimer's disease.

Do I need single-cell RNA-seq data for CapyBio's analysis, or can I use bulk RNA-seq?

Our analyses require single-cell RNA sequencing (scRNA-seq) data. If you don't have scRNA-seq data for your samples yet, we can generate it as an add-on service. Bulk RNA-seq can only be used to build benchmarking references for Capybara analysis.

How is my data used and protected?

Your data is used for one purpose: analyzing your samples and delivering your results. We do not use customer data to train, tune, or improve our algorithms or reference data, and we do not combine it with other customers' data or share it with third parties. Only the CapyBio team members working on your project can access it, and it is handled confidentially throughout the engagement. Because these are services, you own all the results from your data analysis. The only thing we don't disclose is our analysis process, which is CapyBio intellectual property. We also have an MSA template that outlines the key terms for our customers.

Services, packages, and timelines

What analysis packages does CapyBio offer, and what is included in each?

We offer three standard packages:

Analysis packages
PackageWhat it includesMax samplesTurnaround
Capybara™ Benchmarking AnalysisCell type classification, UMAP, differential composition, multi-ID analysis, identity score distributionsUp to 2 samples (50k cells per sample)1 to 2 weeks
Capybara Advanced AnalysisEverything in Benchmarking, plus differentially expressed genes (DEG) and gene set enrichment analysis (GSEA)Up to 8 samples (50k cells per sample)2 to 3 weeks
CellOracle™ Differentiation Optimization AnalysisGene regulatory network construction, per-cluster network centrality, in silico knockout and over-expression of 20 key transcription factorsUp to 2 samples (20k cells per sample)2 to 4 weeks

Every package includes a 1-hour kickoff meeting, an integrated report, and a 1-hour results consultation.

How long does a CapyBio analysis take?

  • Capybara Benchmarking Analysis: 1 to 2 weeks

  • Capybara Advanced Analysis: 2 to 3 weeks

  • CellOracle Differentiation Optimization Analysis: 2 to 4 weeks

Add-ons such as pre-processing, sequencing, or additional samples may extend these timelines.

What deliverables will I receive from a CapyBio analysis?

Every package delivers one integrated report plus a live results consultation. What's in the report depends on the package:

  • Benchmarking: cell type classification, UMAP (top 10 cell types), differential composition analysis, multi-ID analysis, identity score distributions, and an executive summary with next-step recommendations.
  • Advanced: everything above, plus DEG and GSEA.
  • CellOracle: per-cluster network centrality (degree and eigenvector), pairwise cluster comparisons, in silico knockout and over-expression results for 20 transcription factors, and an executive summary with next-step recommendations.

See examples of our data reports.

What if I have more cells or samples than the standard packages include?

Add-on modules extend both platforms:

  • Additional Capybara Sample: extends Benchmarking beyond 2 samples, or Advanced beyond 8 samples.
  • Additional CellOracle Sample: adds samples beyond the 2 included in Differentiation Optimization Analysis.

Does CapyBio offer custom analysis packages?

Yes, we're happy to put together a custom package to fit your needs. Fill out our contact form and someone will be in touch soon.

Does CapyBio offer wet lab services, like scRNA-seq library prep or cell culture?

Yes. In addition to analysis, we offer wet lab services so you can hand off as much of the workflow as you need:

  • Single-cell RNA-seq: library prep, QC, sequencing, and counts-matrix processing, available as an add-on to any package, so you don't need to generate scRNA-seq data elsewhere.
  • Upstream cell culture: cell culture work before sequencing, which may include differentiation and treatment of cells (for example, with compounds or other experimental conditions).
  • Downstream cell culture: follow-up cell culture work after sequencing and analysis, which may also include differentiation and treatment.

We also offer a Pre-Processing, Normalization & Clustering add-on to prepare existing data for CellOracle analysis. Tell us about your project and we'll put together a scope that fits.

Does CapyBio sell differentiated cells?

Coming soon! We have several cell lines and media supplements in development to give scientists the best possible cells for their research. Visit our product page to learn more, and reach out if you have suggestions for products you'd like us to develop.

Use cases

How do I know if my iPSC-derived cells are mature enough for toxicity or efficacy testing?

Many iPSC-derived cells look more like fetal cells than adult ones. iPSC-derived hepatocyte-like cells, for example, often still express fetal markers such as alpha-fetoprotein and show much lower drug-metabolizing enzyme activity than adult primary hepatocytes. A handful of markers can't tell you how far along the maturation path your cells actually are.

Capybara scores each cell against primary human reference data across the whole transcriptome, so you get a quantitative read on how closely your cells match the target identity and how many fall short or sit in transitional states. That gives your team an objective basis for deciding whether a model is fit for purpose before you generate toxicity or efficacy data with it.

Your reference is chosen together with our team. For maturity questions, we usually recommend a reference that spans the differentiation timeline, from iPSC through progenitor stages to the mature cell type, because a more comprehensive reference makes classification more accurate.

What percentage of my differentiated culture is actually my target cell type?

Capybara can tell you. Most differentiation protocols produce a mixed population in which the target cell type is only a subset. Capybara breaks your culture down cell by cell, reporting the fraction that confidently matches the target identity, the fraction matching off-target identities, and the fraction in hybrid or unclassified states.

The differential composition analysis included in every package then shows how those proportions shift between samples or conditions, so you can see whether a protocol change actually improved purity.

Why do some iPSC lines or differentiation batches work better than others, and can CapyBio help me compare them?

Line-to-line and batch-to-batch variability is well documented. In one large study of dopaminergic neuron differentiation, roughly a quarter of iPSC lines consistently failed to produce neurons at all.

Capybara Advanced Analysis benchmarks up to 8 samples side by side (different lines, batches, protocols, or reagent lots) against the same primary reference. Identity score distributions and composition show which samples deviate and by how much, and DEG and GSEA highlight the genes and pathways that separate stronger and weaker performers.

How can I figure out which transcription factors to change to improve my differentiation protocol?

CellOracle Differentiation Optimization Analysis is built for this. It constructs a gene regulatory network from your data and runs in silico knockout and over-expression of 20 key transcription factors, ranking the perturbations predicted to push your cells toward the target fate. It pairs well with Capybara: Capybara shows where your cells stand today, and CellOracle suggests what to change.

My cells pass marker staining and qPCR. Do I still need single-cell analysis to confirm their identity?

Most likely, yes. Markers are necessary but not sufficient. A few genes can be expressed by cells that differ substantially everywhere else in the transcriptome. Markers generally can't distinguish immature from mature cells or reveal partially reprogrammed and hybrid cells, and population-average measurements hide heterogeneity. Researchers have increasingly pointed out these limits of marker-based identity checks.

Capybara uses the whole transcriptome at single-cell resolution, so it captures identity, purity, and off-target populations that a marker panel can miss. It complements functional assays (for example, CYP activity in hepatocytes) rather than replacing them: identity data helps explain why a functional readout looks the way it does.

I develop culture media, reagents, or differentiation supplements. Can CapyBio help me show that my product produces better cells?

Yes. Cells grown with your product and with a standard or competing product can be benchmarked against the same primary human reference, giving you a quantitative, head-to-head comparison of identity scores, purity, and off-target composition. Advanced Analysis supports multiple conditions and adds DEG and GSEA to show which genes and pathways your product shifts. These data can inform your internal product development decisions. You own the results from your analysis, and our MSA template outlines the key terms for our customers.

How close are my iPSC-derived or engineered cells, including cell therapy products, to their primary counterparts?

This is the core question Capybara was designed to answer, whether your cells are iPSC-derived, directly reprogrammed, or expanded and engineered ex vivo. For cell therapy developers, a quantitative measure of identity relative to primary cells can support process development and comparisons across manufacturing runs.

The platform isn't limited to particular cell types, because references are built around the identity you're trying to match. Contact us to talk through your target cell type.