# MiraTyper by MiraOmics

> Cell typing and malignant calls for single-cell, spatial and H&E data.

MiraTyper annotates RNA data with 425 cell types on the Cell Ontology. Single-cell RNA input is AnnData (.h5ad); outputs include cell types, confidence scores, and malignancy probabilities. H&E uses a separate image-based workflow; the 425-type RNA label set is not a claim of 425 types resolved from H&E.

Models and validation depend on the assay and panel. Technologies described on the website include 10x Chromium, Xenium, CosMx, Atera, Singular G4X, Visium, Visium HD, and H&E. Takara Seeker support is a prototype.

Training uses a curated subset of a collection of approximately 100 million cells. The collection size is not the number of cells used for training. Consult the technical details for measured benchmarks, evaluation settings, and limitations.

Inference is delivered in Docker and can run inside the researcher's environment. Transcriptomic inference runs on CPU; H&E and image-based segmentation require GPU resources.

## Product and evidence

- [MiraTyper and supported technologies](https://miratyper.ai/): Product capabilities and assay-specific support.
- [Technical details](https://miratyper.ai/technical/): Inputs, outputs, benchmarks, deployment, and limitations.
- [Case studies](https://miratyper.ai/case-studies/): Published examples with their individual methods and results.
- [H&E inference in TCGA colon cancer](https://miratyper.ai/case-studies/tcga-he-inference/): Exploratory MSI associations, spatial tissue measurements, and validation limits.

## Company and getting started

- [About MiraOmics](https://miratyper.ai/about/): Leadership and partners.
- [Try MiraTyper free](https://miratyper.ai/#free-trial): Add your name and email, then choose a time to discuss running two datasets for free.
- [Contact](https://miratyper.ai/contact/): Send an email and message if scheduling is inconvenient.