Refnd: Preventing Data Leakage in Relational Datasets
Provides a theoretically grounded, scalable solution for preventing data leakage in relational datasets, benefiting practitioners in biochemical and related domains.
The paper addresses data leakage in relational biochemical datasets by introducing the Relational Generative Process (RGP) formalization and Refnd, a splitting algorithm using HNSW for loglinear time proximity graph computation. Validation on an antimicrobial peptide dataset shows Refnd splits yield lower but more realistic performance than traditional splits.
Machine learning models trained on biochemical data are routinely evaluated using splits that fail to account for relational structure, causing information leakage and over-optimistic performance estimates. Existing splitting methods lack theoretical grounding and scale at best quadratically. We introduce the Relational Generative Process (RGP), a mathematical formalization explaining why relational structure arises in biochemical datasets, and Refnd, a splitting algorithm that leverages a proximity graph computed in loglinear time using Hierarchical Navigable Small World (HNSW). We validate on an antimicrobial peptide dataset, showing that Refnd splits yield lower but more realistic evaluation performance than traditional splits. Refnd is applicable to any dataset arising from an RGP such as protein sequences and structures, small molecules, and nucleotide sequences, and is openly available as a Rust accelerated Python package: pip install refnd.