Scikit-bio: Python for Omics

The "missing middle" for omics data analysis in Python

If you work in computational biology your Python import block probably includes the usual suspects: pandas, numpy, matplotlib, and scikit-learn. A new paper in Nature Methods highlights scikit-bio, a library designed to bridge the gap between general data science and the specific needs of biological omics.

Aton, M., McDonald, D., Cañardo Alastuey, J. et al. Scikit-bio: a fundamental Python library for biological omic data analysis. Nat Methods (2025). https://doi.org/10.1038/s41592-025-02981-z. (Read free: https://rdcu.be/eUPjc).

• Website: https://scikit.bio

• Source code (BSD-3): https://github.com/scikit-bio/scikit-bio.

• On PyPI: https://pypi.org/project/scikit-bio/

• On conda-forge: https://anaconda.org/conda-forge/scikit-bio

• Tutorials: https://github.com/scikit-bio/scikit-bio-tutorials

If you’re reading my newsletter you know that omic data (whether genomics, transcriptomics, or metagenomics) presents unique challenges. The authors identify the persistent analytical challenges we face:

high dimensionality (many more features than samples), sparsity (most features are zero) and compositionality (features are interdependent within a sample)

Because of these traits, generic data analysis methods can be inadequate. Scikit-bio addresses these specific pain points with over 500 public-facing functions. Here are a few highlights from the paper:

• Sparsity handling: The library integrates the Biological Observation Matrix (BIOM) format, which is essential for handling sparse tabular data efficiently.

• Compositionality: For dealing with relative abundances scikit-bio implements transformations like center log-ratio (CLR) and differential abundance tests like ANCOM to handle compositional data correctly.

• Tree structures: This is a major strength. The library allows you to construct and manipulate phylogenetic trees and, crucially, integrate them into community modeling. This enables diversity metrics that account for biological relatedness.

Figure 1 from Aton et al 2025: Squared-off boxes with illustrations represent data structures (light grey background) or analyses (white background). Round-cornered boxes represent specific analytical tools, categorized as follows: green, functionality implemented in scikit-bio; blue, functionality offered by external Python libraries that can be used in conjunction with scikit-bio within the Python framework—for example, a distance matrix generated by scikit-bio can be input into SciPy for hierarchical clustering, scikit-learn for k–nearest neighbors classification or umap-learn for UMAP embedding; yellow, functionality provided by non-Python programs that can interact with scikit-bio through file input and output (I/O). For example, scikit-bio can read a phylogenetic tree built by RAxML or a multiple sequence alignment generated by MAFFT.

This library plays nice with the tools you probably already use. It uses numpy arrays and Pandas dataframes.¹ You can compute a distance matrix in scikit-bio and pipe it directly into scikit-learn for clustering or statsmodels for hypothesis testing.

Scikit-bio might sound familiar because it powers much of the functionality in QIIME. However, while it originated in microbiome research, its modular design is now being applied to single-cell data, metabolomics, and epigenomics.

1. This morning while listening to the latest episode of the Test Set podcast from Posit I learned about the Narwhals package, which is a lightweight and extensible compatibility layer between dataframe libraries.