CyteType: AI Cell Annotation for Drug Discovery

Intuitive single cell RNA-seq data analysis and exploration on Nygen Analytics

Annotation isn’t the end point, it’s the bottleneck

Single-cell datasets in biopharma now routinely contain millions of cells across complex disease models. Yet annotation remains constrained by static reference atlases and marker-based approaches that fail to capture disease-relevant biology, creating critical bottlenecks:

"Poor annotation not only slows you down but sends you in the wrong direction. Misidentified cell states become false leads, consuming validation budget and delaying meaningful progress."

Confident cell annotations with explanation & confidence score

Disease & pathway impact summaries at cluster or gene-set level

Literature distillation scans multiple public resources for corroborating evidence in seconds, saving up to 24 weeks of manual review

Current tools force an impossible choice: invest heavily in custom reference building, or accept generic annotations that miss the disease-relevant biology driving your program.

Key Features

Discover the power of Nygen Analytics

Explore how our platform transforms your research with efficient data analysis, enhanced collaboration, and integrated solutions.

Streamlined Data Analysis

Nygen Analytics is an intuitive, visual platform designed for single cell multi omics analysis, enabling scientists to precisely calibrate their data analysis and manage the entire data lifecycle as projects progress and evolve.

Effortless Collaboration

Efficiently collaborate with your bioinformaticians by utilizing flexible data import and export options, allowing you to drive biology-focused analysis while freeing up valuable time for your bioinformatics team.

Effortless Data Hosting

Publish your analysis results in just a few clicks, increasing the visibility of your research. Nygen eliminates the hassle of setting up custom web servers, saving you valuable time and effort.

Built for Biopharma Workflows

Seamless Integration. No Pipeline Rebuilds Required. CyteType fits directly into your existing single-cell analysis pipeline.

AnnData & Seurat Compatible
Integrates within Scanpy or Seurat workflows with minimal setup. Results flow immediately into downstream analyses.

Interactive HTML Reports
Export results as explorable reports with chat-based investigation. Re-annotate clusters with human feedback while maintaining annotation history.

Scales to Drug Discovery Datasets
Tested on millions of cells across 10,000+ clusters. Processes hundreds of clusters per batch with confidence scoring.

Enterprise-Ready Deployment
Multi-tenant architecture ensures data isolation. White-labeled reports support client delivery. API access enables workflow automation.

CyteType (for Scanpy)

CyteTypeR (for Seurat)

Interactive Reports That Unlock Deeper Biology

CyteType generates an interactive HTML report, serving as an exploratory workspace with annotations, evidence, and confidence scores for each cluster.

Chat with the LLM and query your annotated scRNA-Seq clusters

Query at the Cluster Level

CyteType's integrated language models can be queried directly within the report. Users can chat at the cluster level to ask:

→ Why a specific cell type was assigned
→ What markers drove the decision
→ How it relates to known ontologies and reference populations

Interrogate Deeper Biology

Ask the LLM to clarify expression patterns, regulators, or pathway context, linking results to biological mechanisms for deeper insights.

Literature-Linked Reasoning

The model provides cited studies and summaries to support its explanations, ensuring all claims are backed by published evidence.

Interrogate Deeper Biology

Users can refine hypotheses, re-annotate clusters, or create summaries directly within the report, blending machine and human insights.

The Result

A conversational interface over single-cell data that accelerates understanding from clusters to biology. Teams can document reasoning, share insights, and keep an auditable record of decisions.

Frequently asked questions

Everything you need to know about CyteType

How does CyteType differ from SingleR and CellTypist?

SingleR uses correlation-based reference mapping. CellTypist applies logistic regression on curated marker genes. Both rely on pre-trained references and force assignments when cells don't match training data.
‍
CyteType uses multi-agent architecture: separate agents analyze full expression matrices, validate against literature, assess pathway context, and assign confidence scores. The Reviewer Agent rejects weak hypotheses and triggers re-annotation when evidence is insufficient. This enables characterization of populations outside reference atlases through mechanistic context rather than similarity matching.

How are confidence scores calculated?

Confidence scores reflect agreement across multiple evidence sources: marker gene expression strength, reference mapping quality, pathway consistency, and literature support. The Reviewer Agent synthesizes these inputs and assigns scores. Low confidence (<0.5 threshold) triggers re-annotation with alternative hypotheses.

What species are supported?

Human and mouse are fully supported with comprehensive reference atlases and literature coverage. Other species depend on ortholog mapping availability and reference dataset quality. Contact for specific organism requirements.

What happens if I disagree with an annotation?

Interactive reports allow manual override with annotation history tracking. Users can query the integrated LLM to understand reasoning, provide alternative labels, and trigger cluster-level re-annotation. All changes are logged for audit purposes.

Can CyteType annotate spatial transcriptomics data?

Current deployment focuses on dissociated single-cell and single-nucleus RNA-seq. Spatial transcriptomics support is under development.

What if my cell types aren't in standard databases?

The Annotator Agent analyzes full expression patterns and pathway activity independent of reference matching. For populations with weak reference correlation, specialized subagents provide:
‍
- Pathway enrichment (immune signatures, metabolic states)
- Lineage inference (progenitor vs. differentiated markers)
- Disease-context mapping when applicable

Low-confidence annotations include mechanistic explanations rather than forced labels, enabling functional characterization of novel states.

Can CyteType handle disease or perturbed samples?

Yes. Specialized subagents analyze pathway dysregulation, immune exhaustion signatures, and metabolic reprogramming. Disease-context mapping connects observed states to published disease models. CyteType provides both cell identity and functional state annotations, enabling interpretation of pathological phenotypes that don't exist in healthy references.

How do I integrate CyteType into existing pipelines?

CyteType accepts AnnData (Python) or Seurat (R) objects. Returns annotated objects with added metadata fields: cell type assignments, confidence scores, and evidence summaries. API access available for programmatic workflows. No reformatting required.

How frequently are literature databases updated?

PubMed index receives monthly updates. Cell Ontology syncs with quarterly official releases. Model improvements deploy automatically to licensed users.

What data privacy measures are implemented?

Multi-tenant cloud architecture with data isolation. Encryption at rest and in transit. SOC 2 Type II compliant infrastructure. On-premise deployment available for organizations with data residency requirements.

Multi-Agent AI for
scRNA-Seq Cell Type Annotation and Deep Characterisation

Nygen Analytics

CyteType delivers precise, confident cell type identification through multi-layered analysis, moving beyond basic annotation to deep characterisation of cellular identity.

70%Faster

Evidence-Backed

Mechanistic Depth

Annotation isn’t the end point, it’s the bottleneck

Agentic workflow that explains cell states, not just labels

Annotator Agent

Corroborator Agent

Reviewer Agent

Specialized Subagents