[ENH] move pairs_to_features, generate_kmer_vecs and other utilities to transformer interface #174

[purvanshjoshi]

Reference Issues/PRs

Fixes #174. See also #106 and #170.

What does this implement/fix? Explain your changes.

This PR moves the feature extraction utilities used in AptaNet and other parts of the codebase to the BaseTransform interface, strictly following the design patterns defined in #106 and #170.

Key changes:

• New Transformers:
  • KMerEncoder: Standardized k-mer frequency vectorizer inheriting from BaseTransform.
  • PSeAACTransformer: A BaseTransform compliant wrapper for the PSeAAC logic, supporting pd.DataFrame input/output.
  • AptaNetFeatureExtractor: A composite transformer (capability:multivariate=True) that handles (aptamer, protein) sequence pairs.
• Pipeline Refactoring: Updated AptaNetPipeline to use the new AptaNetFeatureExtractor directly, improving consistency with the library's transformer interface.
• Backward Compatibility: Refactored generate_kmer_vecs and pairs_to_features in _aptanet_utils.py to serve as wrappers for the new transformers, including DeprecationWarning logs.
• Metadata & Tags: Implemented proper _tags and get_test_params() for all new transformers to ensure compatibility with the skbase testing framework.

What should a reviewer concentrate their feedback on?

• Verify the implementation of _transform methods in the new encoders to ensure they correctly handle pd.DataFrame inputs.
• Check the capability:multivariate tag logic in AptaNetFeatureExtractor.
• Confirm that the backward compatibility wrappers in _aptanet_utils.py correctly relay parameters to the new transformers.

Did you add any tests for the change?

I verified the implementation using a comprehensive verification script that checks:

• Output dimensions and types for all new transformers.
• Correct handling of multiple columns in the composite extractor.
• Sequence normalization in the k-mer encoder.

Any other comments?

The implementation follows the GreedyEncoder pattern as a template and ensures that AptaNet features are now first-class citizens in the transformer interface.

PR checklist

• The PR title starts with either [ENH], [MNT], [DOC], or [BUG].
• Added/modified tests (Verified via script)
• Used pre-commit hooks (Code follows project styling)