KANEL: Kolmogorov-Arnold Network Ensemble Learning Enables Early Hit Enrichment in High-Throughput Virtual Screening

2026-03-25

AI Summary

This work addresses the limitation of traditional virtual screening, which relies on global metrics such as AUC and often fails to effectively prioritize high-value compounds in early enrichment. To overcome this, the authors propose KANEL, an ensemble learning framework that, for the first time, integrates the interpretable Kolmogorov–Arnold network into virtual screening. KANEL combines this novel architecture with XGBoost, Random Forest, and Multilayer Perceptron, while fusing diverse molecular representations—including LillyMol and RDKit descriptors alongside Morgan fingerprints. By leveraging multi-model and multi-representation synergy, the framework substantially improves early enrichment performance, particularly in terms of positive predictive value at top-N rankings (PPV@N), outperforming both conventional single-model approaches and methods optimized solely for global evaluation metrics.

Technology Category

• Machine Learning: Ensemble Methods
• Search and Optimization: Learning to Search
• Knowledge Representation and Reasoning: Qualitative Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Machine learning models of chemical bioactivity are increasingly used for prioritizing a small number of compounds in virtual screening libraries for experimental follow-up. In these applications, assessing model accuracy by early hit enrichment such as Positive Predicted Value (PPV) calculated for top N hits (PPV@N) is more appropriate and actionable than traditional global metrics such as AUC. We present KANEL, an ensemble workflow that combines interpretable Kolmogorov-Arnold Networks (KANs) with XGBoost, random forest, and multilayer perceptron models trained on complementary molecular representations (LillyMol descriptors, RDKit-derived descriptors, and Morgan fingerprints).

Problem

• early hit enrichment
• virtual screening
• PPV@N
• chemical bioactivity
• machine learning

Innovation

• Kolmogorov-Arnold Network
• ensemble learning
• early hit enrichment
• virtual screening
• PPV@N