The clinical stakes are considerable. MRONJ is defined clinically as exposed bone, or bone that can be probed through a fistula, in the maxillofacial region persisting for more than eight weeks in patients without prior radiation therapy or metastatic disease. Although incidence among osteoporosis patients on antiresorptives is typically below 0.2 percent, it rises to several percent or more among cancer patients receiving high-dose intravenous bisphosphonates. The consequences include chronic bone exposure, pain and impaired chewing, and the fear of the condition often forces clinicians to pause antiresorptive therapy around dental procedures, potentially undermining treatment of the underlying bone disease. Despite decades of investigation, no circulating biomarker has achieved reliable clinical adoption for identifying who is at risk.

Previous candidate markers have repeatedly disappointed. C-terminal telopeptide of type I collagen, tartrate-resistant acid phosphatase 5b, dickkopf-related protein 1, and the RANKL-to-osteoprotegerin ratio were all proposed over the years, but meta-analyses and guideline committees have largely rejected them, citing small, heterogeneous, single-centre studies and a hypothesis-driven focus on a narrow set of bone-turnover proteins. The Korean team, led by researchers at Seoul National University and Ewha Womans University, took a deliberately different route: an unbiased, high-resolution survey of the entire serum proteome, followed by validation in separate cohorts and population-scale data.

In the discovery phase, the researchers applied tandem mass tag-based quantitative proteomics to serum from 25 MRONJ patients and 16 medication-exposed controls, depleting the fourteen most abundant serum proteins to reveal the deeper proteome. The workflow quantified 1,481 proteins with high technical reproducibility, with median coefficients of variation between internal standard channels of roughly four percent. Among all of these, one protein stood out: dopamine β-hydroxylase, or DBH, was significantly less abundant in the MRONJ group, with an adjusted P value of 0.037 after Benjamini–Hochberg correction for multiple testing.

DBH is the enzyme that converts dopamine into norepinephrine inside sympathetic neurons and adrenal chromaffin cells, and it is co-released with norepinephrine during sympathetic activation. Because it circulates more stably than norepinephrine itself, it has long been proposed as a surrogate window onto sympathetic tone, though earlier small studies produced inconsistent results. To confirm the proteomic signal, the team measured DBH by enzyme-linked immunosorbent assay and found excellent agreement with the mass spectrometry data, with a Pearson correlation coefficient of 0.962. Median serum DBH in MRONJ patients was 915.1 picograms per millilitre, compared with 1,317 picograms per millilitre in controls, a difference that was highly statistically significant.

The association survived rigorous statistical scrutiny. In multivariable logistic regression adjusted for age and sex, each one-standard-deviation decrease in DBH was associated with nearly twelve-fold higher odds of MRONJ in the discovery cohort, with a receiver operating characteristic area under the curve of 0.848 and an optimal cut-off of 1,082 picograms per millilitre yielding 84 percent sensitivity and 81.3 percent specificity. Crucially, an independent validation cohort of 55 additional participants reproduced the direction of the effect, with lower DBH again significantly associated with MRONJ after adjustment. Sensitivity analyses excluding cancer patients, anti-angiogenic drug users, beta-blocker users, and users of any sympathetically active medication all preserved the association, arguing against confounding by concurrent drugs or comorbidities.

Population-scale data provided a third, independent line of evidence. Mining plasma proteomic measurements from 44,740 UK Biobank participants of white British ancestry, the researchers narrowed the analysis to 2,270 people with osteoporosis or bisphosphonate exposure, identifying 10 MRONJ cases and 2,224 controls. Using Firth’s penalized likelihood logistic regression to handle the rare outcome, they found that each standard-deviation decrease in plasma DBH was associated with 1.57-fold higher odds of MRONJ after adjusting for age, sex, body mass index, bisphosphonate use and genetic ancestry. The prevalence of MRONJ in this subset, 0.45 percent, matched expectations for antiresorptive-exposed populations, lending credibility to the case definitions.

To establish that circulating DBH genuinely tracks sympathetic activity, the team turned to blood pressure, a classic downstream readout of sympathetic tone. Across 34,948 UK Biobank participants, plasma DBH was positively associated with both systolic and diastolic blood pressure after adjustment for age, sex, body mass index, smoking, drinking and genetic principal components. Spline and quantile regression analyses showed the relationship strengthening at higher blood pressures, and participants in the top 30 percent of DBH levels had significantly elevated odds of hypertension. These data rehabilitate DBH as a population-level marker of sympathetic tone and suggest that the low DBH seen in MRONJ patients reflects genuinely altered sympathetic regulation rather than a mere biochemical curiosity.

The mechanistic twist came from the laboratory. In a mouse model in which repeated zoledronic acid administration combined with maxillary molar extraction reliably produces jawbone necrosis, the researchers added propranolol, a non-selective beta-adrenergic blocker, to pharmacologically suppress sympathetic signalling. Propranolol did not change bone structure beyond what zoledronic acid achieved, but it dramatically worsened the disease itself: histology revealed significantly larger necrotic bone areas in the combination group than with zoledronic acid alone. Tartrate-resistant acid phosphatase staining showed that propranolol further reduced the number of giant, dysfunctional osteoclasts that zoledronic acid induces, particularly at non-extraction sites, indicating an additive suppression of bone resorption on top of an already profound antiresorptive state.

Co-culture experiments fleshed out the cellular logic. When primary mouse osteoblasts were stimulated with the beta-adrenergic agonist isoproterenol, the ratio of RANKL to osteoprotegerin, the key osteoblast-derived signals that drive osteoclast differentiation, rose sharply, and propranolol abolished this effect. Zoledronic acid left the RANKL-to-osteoprotegerin ratio untouched, yet in functional osteoblast–osteoclast co-cultures the two drugs acted through distinct pathways to inhibit osteoclast formation additively. Since excessive suppression of bone resorption is widely regarded as the central mechanism of MRONJ, the authors propose that reduced sympathetic signalling, as reflected by low circulating DBH, may push osteoclast activity below the threshold needed for jawbone maintenance, particularly in the high-turnover alveolar bone of the jaw. The team is careful to note limitations: DBH was measured after disease onset, so prospective studies are needed to confirm predictive value, and the UK Biobank case definitions carry some classification uncertainty. Even so, the convergence of clinical proteomics, population genetics-scale validation and animal experimentation makes circulating DBH the most credible candidate biomarker for MRONJ risk stratification to date, and opens an entirely new chapter in the neuro-skeletal biology of the jaw.

Subject of Research: Circulating dopamine β-hydroxylase as a biomarker and sympathetic dysregulation mechanism in medication-related osteonecrosis of the jaw

Article Title: Decreased circulating dopamine β-hydroxylase is associated with medication-related osteonecrosis of the jaw and implicates sympathetic dysregulation

Article References: Lee, K., Cho, H., Kim, J. Y., Hwang, H., Song, J.-S., Yoon, S., Jung, S., Kang, Y.-H., Kim, Y. H., Kim, J.-Y., Kim, S.-J., Park, H., & Kim, J.-W. (2026). Decreased circulating dopamine β-hydroxylase is associated with medication-related osteonecrosis of the jaw and implicates sympathetic dysregulation. Experimental & Molecular Medicine. https://doi.org/10.1038/s12276-026-01868-z

Image Credits: AI Generated

DOI: 10.1038/s12276-026-01868-z

Keywords: medication-related osteonecrosis of the jaw, dopamine beta-hydroxylase, sympathetic nervous system, bisphosphonates, antiresorptive therapy, serum proteomics, UK Biobank, osteoclasts, biomarker, propranolol, bone remodeling, norepinephrine

The work, led by Ashwin Srinivasan of BITS Pilani in Goa, together with Tirtharaj Dash, A. Baskar, Michael Bain, Sanjay Kumar Dey, and Mainak Banerjee, addresses a class of hybrid neurosymbolic models that remains relatively under-explored. Rather than letting a neural generator run free and filtering its output afterward, SNG inverts the relationship. A symbolic learner first examines logical specifications of what counts as feasible data, sometimes from as few as a single example. That specification then constrains the conditional information supplied to a neural generator, which is required to reject any candidate instance that violates the symbolic description. The result is a pair consisting of a symbolic description of feasible instances and a set of newly generated instances that provably satisfy it.

Technically, the framework rests on a formal semantics built from partially ordered sets. The researchers construct what they describe as base and fibre posets, combined into an overall partial order that characterizes the space of hybrid systems. The symbolic component in their implementation draws on a restricted form of Inductive Logic Programming, a family of techniques dating back decades that learns logical rules from examples and background knowledge. The neural component is a large language model prompted to produce candidate molecules as SMILES strings, the standard text encoding of molecular structure. Each candidate is checked against the learned logical specification before it is accepted, ensuring that every generated molecule conforms to the constraints the symbolic learner has inferred.

To score candidate hypotheses, the system employs a Bayesian heuristic originally developed in the ILP literature, which balances the prior probability of a hypothesis against how well it explains positive and negative examples. The search over hypotheses proceeds through nested hyper-rectangles defined over molecular factors such as predicted binding affinity, molecular weight, and synthetic accessibility score. Sampling within these rectangles uses a Latin Hypercube approach, and each successive experiment narrows the constraints, producing hypotheses that are logically entailed by their predecessors. The authors prove formal properties about this procedure, including guarantees that the generated sets remain within the feasible extension of the learned hypothesis.

The application that gives the work its urgency is early-stage drug design, specifically the discovery of lead compounds, the validated starting points from which medicinal chemists develop actual drug candidates. The team evaluated their SNG implementation on benchmark protein targets whose biology is well understood, including JAK2 and DRD2, alongside dopamine beta-hydroxylase, or DBH, a comparatively poorly studied enzyme implicated in cardiovascular disease. Binding affinities of generated molecules were estimated computationally using the GNINA molecular docking program, which predicts how strongly a small molecule will lodge itself in the binding pocket of a protein target.

The results are striking. On the benchmark problems, where drug targets are well characterized, SNG performance was statistically comparable to state-of-the-art methods for molecular generation. More intriguing still were the exploratory experiments on DBH, a target for which little domain knowledge exists in the literature. Molecules generated by the system for this target exhibited predicted binding affinities on par with leading clinical candidates. Because the symbolic learner had to infer target-specific constraints from just five known inhibitors, the achievement suggests the framework can operate productively even in the data-sparse regimes that characterize most real drug discovery projects, where high-throughput screening data is expensive and incomplete.

The choice of neural generator turned out to matter in unexpected ways. The researchers compared OpenAI’s GPT-4o, Anthropic’s Claude 3.5 Sonnet, and a small 87-million-parameter GPT-2 model trained on roughly 480 million molecules from the ZINC chemical database. The general-purpose models performed well on heavily studied targets like JAK2 and DRD2, presumably because their vast training corpora include relevant biochemical literature. But the chemistry-specialized Molecule-GPT2 outperformed them on DBH, the lesser-studied target, where knowledge had to be extracted from the inhibitor molecules themselves rather than recalled from pretraining. Notably, the entire study cost only around ten US dollars in API fees per commercial model, underscoring that the framework is not wedded to any particular language model and could accommodate open-source alternatives.

Perhaps the most persuasive endorsement came from human experts. A structural biologist and a synthetic organic chemist reviewed the system’s outputs and found the symbolic specifications genuinely useful as preliminary filters for triaging candidates. Several of the generated molecules were judged viable for synthesis and wet-laboratory testing, a meaningful threshold for any computational design method. The experts’ assessments also highlighted a practical advantage of the neurosymbolic approach: because the system produces an explicit logical description of what makes a molecule feasible, chemists can inspect, critique, and refine the criteria rather than treating the generator as an inscrutable black box.

The authors are careful to situate their contribution within the broader neurosymbolic movement, sometimes called the third wave of neurosymbolic AI, which seeks to combine learning and reasoning rather than choosing between them. They acknowledge that the categorization of systems into purely symbolic or purely neural roles is necessarily a simplification, since induction typically requires deduction and complex reasoning involves trial-and-error learning. They also conjecture that the symbolic learner at the heart of SNG may yield more reliable hypotheses and reduce the number of experiments a laboratory robot would need to conduct in automated discovery loops. Extensions to wider classes of hybrid systems, including those where the neural component plays the primary role, are outlined in the paper’s appendices.

The implications extend well beyond one enzyme or one pharmaceutical project. By demonstrating that logical specifications learned from tiny datasets can steer a generative language model toward chemically meaningful and synthetically plausible outputs, the study offers a template for constrained generation in any scientific domain where correctness matters, from materials design to protein engineering. The researchers have released their code and data openly through a public repository, and the framework’s modest computational demands mean that laboratories without access to frontier-scale models could adopt the approach immediately. As generative AI moves from producing text to proposing tangible scientific artifacts, frameworks like SNG suggest that the path forward lies not in bigger models alone, but in models that know the rules they must obey.

Subject of Research: Hybrid neurosymbolic AI systems that integrate symbolic learning with neural generation for discovering lead molecules in early-stage drug design

Article Title: Symbolic Neural Generation with Applications to Lead Discovery in Drug Design

Article References: Srinivasan, A., Dash, T., Baskar, A., Bain, M., Dey, S. K., & Banerjee, M. (2026). Symbolic Neural Generation with Applications to Lead Discovery in Drug Design. Machine Learning, 115(9), Article 210. https://doi.org/10.1007/s10994-026-07136-5

Image Credits: AI Generated

DOI: 10.1007/s10994-026-07136-5

Keywords: neurosymbolic AI, drug discovery, large language models, inductive logic programming, molecular generation, lead discovery, machine learning, binding affinity, symbolic reasoning, drug design, GPT-4o, dopamine beta-hydroxylase