Supported modalities

As of 1 Oct 2025, only the modalities below are supported by K-Pro "Bring Your Own Data":

• Clinical data

• Molecular data:

  • Bulk RNA seq (TPM, FPKM, DESeq2)

  • Single cell RNA seq (unnormalized, SCTransform-corrected counts, Counts per million.)

  • Spatial transcriptomics (SCTransform, CPM, SME Normalization)

  • Whole Exome Sequencing (WES)

• Histology - slide annotation linked to a patient

Clinical data

Standard clinical data formats are supported (.csv, .tsv, .xlsx, …).

Molecular data

Here is a table of supported modalities and source formats.

Modalities	Formats
Bulk RNAseq	Count matrix (.txt / .tsv / .csv); AnnData (.h5ad)
Single cell RNAseq	Matrix Market (.mtx + .tsv); HDF5 (.h5); AnnData (.h5ad) Seurat object (.rds)
Spatial transcriptomics	Matrix Market (.mtx + barcodes.tsv + features.tsv); HDF5 (.h5); AnnData (.h5ad) Seurat object (.rds)
WES / WGS	VCF

Note that Owkin also developed an in-house processing pipeline that can process FASTQ files for both bulk RNAseq, Chromium single-cell and Visium spatial sequencing.

Imaging data

Here is a table of supported modalities and formats.

Modalities	Formats
Histology (H&E WSI)	.tif, .tiff, .svs, .dcm, .svs, .ndpi, .mrxs
Immunohistochemistry (IHC)	.tif, .tiff, .svs, .dcm, .svs, .ndpi, .mrxs

Note that Owkin developed several imaging processing pipelines that can be applied to imaging data, including:

• Cell segmentation and cell annotation via the HIPE model. Resulting cell-type quantifications are then saved (format: .csv) and used by K pro.

• IHC score extraction with proprietary models (e.g., HER2 score, NMR, etc.). Resulting patient-level scores are then saved and used by K-pro.

• Self Supervised Features extraction via Foundation Models: slides are divided into tiles from which latent representations are computed, saved (format: .npy) and used by K pro. Our models rank amongst the best on various downstream tasks (see XX)