Single Cluster Analysis

Overview

The runCASSIA function analyzes a single cluster of marker genes to identify the cell type. This function is specifically designed for users who only have one cluster to analyze.

Note: CASSIA is designed to handle multiple clusters at once via Batch Processing. Use this function when you need to annotate just a single cluster.

Quick Start

result <- runCASSIA(
    marker_list = c("CD3D", "CD3E", "CD2", "TRAC"),
    model = "anthropic/claude-sonnet-4.5",
    tissue = "blood",
    species = "human",
    provider = "openrouter"
)

# View the annotation result
print(result$structured_output)
R

For model recommendations, see How to Select Models and Providers.

Input

Marker List Format

Provide a character vector of marker gene names for your cluster:

marker_list <- c("CD3D", "CD3E", "CD2", "TRAC", "IL7R")
R

These should be the top differentially expressed genes that characterize your cluster of interest.

Parameters

Required

Parameter	Description
`marker_list`	Character vector of marker gene names for the cluster
`model`	LLM model ID (e.g., `"anthropic/claude-sonnet-4.5"`)
`tissue`	Tissue type (e.g., `"blood"`, `"brain"`)
`species`	Species (e.g., `"human"`, `"mouse"`)
`provider`	API provider (`"openrouter"`, `"openai"`, `"anthropic"`)

Optional

Parameter	Default	Description
`temperature`	0	Output randomness (0=deterministic, 1=creative). Keep at 0 for reproducible results.
`additional_info`	`NULL`	Extra experimental context about the sample
`validator_involvement`	`"v1"`	Validation intensity: `"v1"` (moderate) or `"v0"` (high, slower)
`reasoning`	`NULL`	Reasoning depth for compatible models (`"low"`, `"medium"`, `"high"`). See below.

Parameter Details

Model Selection

Default: anthropic/claude-sonnet-4.5 for best performance
Alternative: google/gemini-2.5-flash for faster analysis
When using OpenRouter, specify the complete model ID
See How to Select Models and Providers for detailed recommendations

Reasoning Parameter

Controls reasoning depth for compatible models (GPT-5 series via OpenRouter)
Options: "low", "medium", "high"
Omit this parameter for standard mode
See Reasoning Effort Parameter for details

Additional Context

Use additional_info to provide experimental context
Example: "Sample from tumor microenvironment, focus on immune infiltration"

Output

The function returns a list with two components:

Component	Description
`structured_output`	The annotation result containing predicted cell type and reasoning
`conversation_history`	Complete conversation log for debugging and transparency