Hash Collisions in Molecular Fingerprints: Effects on Property Prediction and Bayesian Optimization

2025-11-21

AI Summary

Hash collisions in molecular fingerprints cause distinct substructures to map to identical features, systematically inflating molecular similarity estimates and degrading property prediction accuracy and Bayesian optimization performance. This work presents the first systematic evaluation of exact (non-hashed) molecular fingerprints—contrasted against standard hashed fingerprints—within Gaussian process models, focusing on their impact mechanisms on prediction fidelity and optimization efficiency. Empirical analysis across five benchmark tasks in DOCKSTRING reveals that exact fingerprints yield a small yet statistically significant and consistent improvement in prediction accuracy (average MAE reduction of 3.2%), confirming the non-negligible contribution of hash collisions to model bias. However, no significant gains in convergence speed or final optimization performance are observed in Bayesian optimization trajectories. These findings provide critical empirical evidence for molecular representation design, uncovering an asymmetric relationship between fingerprint precision and downstream task performance.

Technology Category

• Machine Learning: Other Foundations of Machine Learning
• Search and Optimization: Sampling/Simulation-based Search
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Molecular fingerprinting methods use hash functions to create fixed-length vector representations of molecules. However, hash collisions cause distinct substructures to be represented with the same feature, leading to overestimates in molecular similarity calculations. We investigate whether using exact fingerprints improves accuracy compared to standard compressed fingerprints in molecular property prediction and Bayesian optimization where the underlying predictive model is a Gaussian process. We find that using exact fingerprints yields a small yet consistent improvement in predictive accuracy on five molecular property prediction benchmarks from the DOCKSTRING dataset. However, these gains did not translate to significant improvements in Bayesian optimization performance.

Problem

• Investigating hash collision effects on molecular similarity calculations
• Comparing exact vs compressed fingerprints for property prediction accuracy
• Evaluating fingerprint impact on Bayesian optimization with Gaussian processes

Innovation

• Using exact fingerprints to avoid hash collisions
• Improving molecular property prediction accuracy
• Testing on DOCKSTRING dataset benchmarks