API

Toolkit functions

Preprocessing

Functions

generate_gene_df(ad, house_keeping_genes[, ...])	Generates a DataFrame with the top 50 genes for each observation in an AnnData object.
segment_patches(img_array, coord, patch_dir)	Extracts small image patches centered at specified coordinates and saves them as individual PNG files.
get_library_id(adata)	Retrieves the library ID from the AnnData object, assuming it contains spatial data.
get_scalefactors(adata[, library_id])	Retrieves the scalefactors from the AnnData object for a given library ID.
get_spot_diameter_in_pixels(adata[, library_id])	Retrieves the spot diameter in pixels from the AnnData object's scalefactors for a given library ID.
prepare_data_for_alignment(data_path[, ...])	Prepares data for alignment by reading an AnnData object and preparing the high-resolution tissue image.
load_data_for_annotation(st_data_path, json_path)	Loads spatial transcriptomics (ST) data from an .h5ad file and prepares it for annotation.

Utilities

Functions

load_model(model_path, device)	Load pretrained OmiCLIP (COCA ViT‑L‑14) model, its image preprocess, and tokenizer.
encode_images(model, preprocess, ...)	Batch–encode a list of image file paths into L2‑normalized embeddings.
encode_texts(model, tokenizer, texts, device)	Batch–encode a list of strings into L2‑normalized embeddings.
encode_text_df(model, tokenizer, df, ...)	Encodes an entire DataFrame column into (N, D) embeddings.

Loki Align

Functions

apply_homography(coordinates, H)	Apply a 3x3 homography matrix to 2D coordinates.
align_tissue(ad_tar_coor, ad_src_coor, ...)	Aligns the source coordinates to the target coordinates using Coherent Point Drift (CPD) registration, and applies a homography transformation to warp the source coordinates accordingly.
find_homography_translation_rotation(...)	Find the homography between two sets of coordinates with only translation and rotation.

Classes

DeformableRegistration([alpha, beta, ...])	Deformable registration.
EMRegistration(X, Y[, sigma2, ...])	Expectation maximization point cloud registration.

Loki Annotate

Functions

annotate_with_bulk(img_features, bulk_features)	Annotates tissue image with similarity scores between image features and bulk RNA-seq features.
annotate_with_marker_genes(classes, ...)	Annotates tissue image with similarity scores between image features and marker gene features.
load_image_annotation(image_path)	Loads an image with annotation.

Loki Decompose

Functions

generate_feature_ad(ad_expr, feature_path[, sc])	Generates an AnnData object with OmiCLIP text or image embeddings.
cell_type_decompose(sc_ad, st_ad[, ...])	Performs cell type decomposition on spatial data (ST or image) with single-cell data .

Loki Retrieve

Functions

retrieve_st_by_image(image_embeddings, ...)

Retrieves the top-k most similar ST based on the similarity between ST embeddings and image embeddings.

Loki PredEx

Functions

predict_st_gene_expr(image_text_similarity, ...)

Predicts ST gene expression by H&E image.

Plotting

Functions

plot_alignment(ad_tar_coor, ad_src_coor, ...)	Optimized plot: target, source, and aligned coordinates with titles.
plot_alignment_with_img(ad_tar_coor, ...[, s])	Optimized plot with images in the background and subplot titles.
plot_img_with_annotation(fullres_img, ...)	Plots image with polygons.
plot_annotation_heatmap(st_ad, roi_polygon, ...)	Plots tissue type annotation heatmap.