Introduction
AAanalysis is a Python framework designed for scientists and researchers focusing on interpretable sequence-based protein prediction. Ideal for comparing protein sequences using amino acid scales, this toolkit is versatile enough for any sequence analysis representable by numerical values.
Key Algorithms
CPP: Comparative Physicochemical Profiling, an interpretable feature engineering algorithm comparing two sets of protein sequences to identify the set of most distinctive features.dPULearn: A deterministic Positive-Unlabeled (PU) Learning algorithm tailored for training on unbalanced and small datasets, enhancing predictive accuracy.AAclust: A k-optimized clustering wrapper that selects redundancy-reduced sets of numerical scales, such as amino acid scales.
Purpose and Audience
For computational biologists, bioinformaticians, and protein engineers, AAanalysis facilitates the analysis and comparison of proteins to discover interpretable physicochemical signatures, the features that distinguish groups of proteins and underlie their biological interactions and functions. These signatures span the whole workflow, from simple sequence analysis to interpretable protein prediction and protein engineering, integrating state-of-the-art explainable AI (XAI) methods.
Overview of Documentation
The documentation is organized into four sections: Overview, Guides, Reference, and Project.
Overview: New to AAanalysis? Begin with Getting Started to install the package and run your first analysis. Delve into the core concepts and design philosophy behind the algorithms in the Usage Principles section, equipping you with the mental models necessary for effective application — including the evaluation strategies for a transparent, objective analysis of the algorithms´ outcomes.
Guides: For hands-on experience, the Tutorials teach each tool with its parameters and outputs, the Protocols walk through complete, end-to-end analyses for biological questions, and the Use Cases showcase published studies end to end from bundled data.
Reference: Look up the exact signatures in the API documentation, or reach for the
one-call golden pipelines in API (Pipelines). Browse the overview
Data Tables (including the AAontology scale classification and benchmark protein
datasets) and the DataFrame schemas that define every df_* contract, check the
Glossary of key terms, and discover the scientific foundation of AAanalysis in the
Scientific References section.
Project: Development conventions and how to contribute live in the Contributing guide, the Docstring Guide documents the docstring style, and the Release Notes track changes across versions.
Finally, four at-a-glance reference documents summarise the whole framework. Keep them open while you work:
Cheat Sheet: the canonical workflow, the main classes by capability, and the Part × Split × Scale feature ontology on three pages.
Decision Map: a flowchart from your goal (explore, predict, or optimize) to the exact AAanalysis class or function to call.
Ecosystem Map: where AAanalysis fits among related bioinformatics, machine-learning, and explainability tools (or read the full positioning article with the map plus background).
Data Flow Map: how data flows from sequences and scales through CPP to features, models, and explanations.