dynamo-preprocess

You can convert an AnnData preprocessing notebook into a reusable agent skill by wrapping it with the dynamo.preprocessing.Preprocessor. This translates notebook-style recipes into deterministic, configurable steps using the preprocess_adata interface.

Preprocessing AnnData for downstream dynamo velocity analyses is best done by selecting a recipe like monocle or seurat and validating essential outputs. The skill verifies normalized layers, obsm X_pca, and adata.obs/var keys to ensure compatibility with velocity workflows.

Yes, you can configure parameters for recipes like sctransform and pearson_residuals using the config_*_recipe tuning interface. This allows you to customize per-recipe settings within the dynamo Preprocessor wrapper for your specific preprocessing needs.

The preprocessing wrapper supports both a wrapper workflow for common cases and a stepwise path. The stepwise path allows you to debug custom AnnData preprocessing pipelines or inspect intermediate stages before reaching the final recipe output.

A dynamo preprocessing recipe may fail to generate the expected obsm X_pca layer due to incorrect recipe selection or parameter configuration. The skill validates essential outputs like obsm X_pca and normalized layers to catch these issues early.

You need to use specific preprocessing recipes like monocle_pearson_residuals over seurat when your AnnData requires particular normalization and variance stabilization. The skill provides access to five recipe branches, allowing you to select the method that best fits your data characteristics.