Hybrid modeling with attribution-guided symbolic enhancements of known physics: Applications to complex biological and engineering systems
- 1Department of Chemical & Biomedical Engineering, FAMU-FSU College of Engineering, Tallahassee, FL, 32310, USA
- 2William G. Lowrie Department of Chemical and Biomolecular Engineering, The Ohio State University, Columbus, OH, 43210, USA
- 3Artie McFerrin Department of Chemical Engineering, Texas A&M University, College Station, TX, 77843, USA
Abstract
Hybrid mechanistic-machine learning models combine first-principles equations with data-driven components to compensate for incomplete system knowledge, promising to bridge mechanistic understanding and data-driven flexibility. In practice, however, model discrepancy is often absorbed into black-box corrections, limiting interpretability and scientific insight. In this work, we propose HATS (Hybrid modeling with ATtribution-guided Symbolic regression), a framework that replaces opaque data-driven corrections with compact, interpretable symbolic representations. HATS iteratively trains a hybrid mechanistic–data-driven model and uses shapley additive explanation (SHAP) based attribution analysis to prune weakly influential inputs, thereby localizing the dominant sources of model mismatch. Symbolic regression is then applied to recover closed-form expressions for the unknown dynamics under explicit complexity constraints. When direct symbolic recovery is infeasible, HATS learns a low-dimensional latent representation of the uncertainty and performs symbolic regression in the latent space, yielding simple and interpretable models. Unlike existing hybrid modeling approaches, HATS adaptively determines where, why , and how uncertainty should be represented, combining feature attribution, symbolic regression, and latent learning within a unified framework. We demonstrate the approach through an in silico rediscovery of the TNF- α signaling pathway and an industrial-scale fermentation case study, thereby highlighting HATS as a novel pathway toward transparent, discovery-oriented hybrid modeling of complex systems.
Related
- Data-Driven Modeling of Ship Maneuvering Dynamics via Control-Aware Hankel-DMDc and Mode Attribution
- Using high pressure torsion to process magnesium alloys for biological applications
- Graphene Layers as Functional Biomaterials. From Fabrication and Characterization to Biological Applications
- Pure FENE Bond Potential for Soft Matter and Biological Simulations: Theory, HOOMD-blue Implementation, and Applications to Polymer, Colloidal, and Membrane Systems
- Pilot search for continuous gravitational waves from Hercules X-1 in O3 LIGO data: enhancements to a hidden Markov model
BibTeX
@article{kudva2026hybrid,
title = {{Hybrid modeling with attribution-guided symbolic enhancements of known physics: Applications to complex biological and engineering systems}},
author = {Akshay Kudva and Hangjun Cho and Parth Shah and Dongheon Lee and Joseph Sang‐Il Kwon},
journal = {Chemical Engineering Science},
year = {2026},
month = may,
doi = {10.1016/j.ces.2026.124269},
url = {https://doi.org/10.1016/j.ces.2026.124269},
note = {CC BY},
}
RIS
TY - JOUR
TI - Hybrid modeling with attribution-guided symbolic enhancements of known physics: Applications to complex biological and engineering systems
AU - Akshay Kudva
AU - Hangjun Cho
AU - Parth Shah
AU - Dongheon Lee
AU - Joseph Sang‐Il Kwon
PY - 2026
DA - 2026/05/22
JO - Chemical Engineering Science
DO - 10.1016/j.ces.2026.124269
UR - https://doi.org/10.1016/j.ces.2026.124269
AB - Hybrid mechanistic-machine learning models combine first-principles equations with data-driven components to compensate for incomplete system knowledge, promising to bridge mechanistic understanding and data-driven flexibility. In practice, however, model discrepancy is often absorbed into black-box corrections, limiting interpretability and scientific insight. In this work, we propose HATS (Hybrid modeling with ATtribution-guided Symbolic regression), a framework that replaces opaque data-driven corrections with compact, interpretable symbolic representations. HATS iteratively trains a hybrid mechanistic–data-driven model and uses shapley additive explanation (SHAP) based attribution analysis to prune weakly influential inputs, thereby localizing the dominant sources of model mismatch. Symbolic regression is then applied to recover closed-form expressions for the unknown dynamics under explicit complexity constraints. When direct symbolic recovery is infeasible, HATS learns a low-dimensional latent representation of the uncertainty and performs symbolic regression in the latent space, yielding simple and interpretable models. Unlike existing hybrid modeling approaches, HATS adaptively determines where, why , and how uncertainty should be represented, combining feature attribution, symbolic regression, and latent learning within a unified framework. We demonstrate the approach through an in silico rediscovery of the TNF- α signaling pathway and an industrial-scale fermentation case study, thereby highlighting HATS as a novel pathway toward transparent, discovery-oriented hybrid modeling of complex systems.
SN - 0009-2509
SN - 1873-4405
ER -