Automated Discovery of Therapeutic Biomaterial for Renally Impaired Hyperuricemia Patients by Natural Language Processing and Machine Learning

Xiaodong Zeng, Jiahao Qiu, Xin Zhao, Kangfei Liu, Longfei Zhao, Junling An, Xiang Liang, Tianzhi Liu, Zhichao Wang, Weidi Xie, Mengdi Wang, Jie Luo, Shiyi ZhangbioRxiv 2025 · DOI

Abstract

The exponential growth of scientific publications presents opportunities for researchers to identify valuable knowledge, especially in the highly interdisciplinary field of biomaterials. However, marrying unmet clinical needs with existing materials reported for other purposes is extremely challenging. Here, we designed a multi-tiered framework, MatWISE, that fuses state-of-the-art natural language processing, semantic relationship mapping, and machine learning to automate material discovery. We successfully identified and optimized $\delta$-MnO₂ into an orally administered, nonabsorbable uric acid (UA) lowering biomaterial. $\delta$-MnO₂ demonstrated superior UA-lowering effects in mouse models compared to standard-of-care drugs and is highly promising as a safe treatment for renally impaired hyperuricemia patients.

Introduction: The Challenge of Hyperuricemia in Renal Impairment

Hyperuricemia is a major risk factor for gout and chronic kidney disease (CKD). Most current uric acid (UA)-lowering drugs, such as allopurinol, are cleared by the kidneys. For patients with impaired renal function, these medications can become toxic or ineffective.

An ideal solution is a "non-systemic" therapy: an oral material that remains in the gut, adsorbs uric acid from the intestinal tract, and carries it out of the body via feces. Finding such a material—safe, insoluble, and with high UA affinity—is a significant challenge given the vast volume of existing scientific literature. With over 20,000 papers published on uric acid alone, human researchers cannot manually evaluate every candidate material.

The MatWISE Discovery Framework

To solve this "needle in a haystack" problem, researchers developed MatWISE (Materials Workflow for Intelligent Screening and Evaluation). This AI-driven pipeline mines scientific literature to uncover "latent knowledge" that human researchers might miss.

The MatWISE AI-driven discovery pipeline and experimental screening process.

The framework operates in several stages:

  1. Information Retrieval & NER: 20,000 abstracts were pulled from Scopus. The Stanza NLP library performed Named Entity Recognition (NER) to identify over 200,000 chemical terms.
  2. Semantic Mapping with BioGPT: Instead of simple keyword searches, the team used BioGPT to generate high-dimensional vector embeddings. Materials were ranked based on their semantic similarity to the concept of "adsorption," identifying candidates that behave as adsorbents even if not previously tested for UA.
  3. LLM Filtering with GPT-4: GPT-4, equipped with web-search capabilities, filtered the top 288 candidates for specific physical properties: solid state, water-insolubility, non-toxicity, and cost-effectiveness (<$3/gram).

Experimental Validation and Therapeutic Mechanism

The AI-driven filtering narrowed the field to 24 materials for laboratory testing. Among these, $\delta$-MnO₂ (birnessite) emerged as the superior candidate, showing a maximum adsorption capacity of 1357.22 mg/g—the highest reported for any UA adsorbent.

Mechanism of oral δ-MnO₂ in the intestine for uric acid lowering.

The study revealed that $\delta$-MnO₂ employs a dual-action mechanism. It doesn't just adsorb uric acid; it also acts as a catalyst, oxidizing UA into allantoin, a more soluble substance that is easier for the body to eliminate. As a non-absorbable material, it passes through the GI tract and is fully eliminated in feces within 8 hours, bypassing the kidneys entirely.

Key Results and Clinical Potential

In vivo testing across three hyperuricemia mouse models confirmed the efficacy of oral $\delta$-MnO₂. Key findings include:

  • Significant UA Reduction: Serum UA levels were lowered more effectively than with standard-of-care drugs.
  • Renal Protection: By diverting UA excretion to the gut, the material reduced the load on the kidneys, potentially preventing kidney stones and further renal strain.
  • Safety Profile: No manganese release into the bloodstream was detected, and the material showed no toxicity over a 14-day study.

Conclusion and Future Directions

The discovery of $\delta$-MnO₂ demonstrates a new research paradigm for biomaterials, combining machine learning with targeted experimentation to address unmet clinical needs. Future work will focus on long-term human safety studies and evaluating the impact on the gut microbiome. The MatWISE framework itself holds potential for adaptation to discover therapeutic materials for various other complex diseases.

Summary updated

Abstract

The exponential growth of scientific publications presents opportunities for researchers to identify valuable knowledge, especially in the highly interdisciplinary field --- biomaterials, where exploiting possible connections between unmet clinical needs and materials properties from literatures is crucial. However, with traditional literature reading, it is extremely challenging to marry unmet clinical needs with existing materials reported for different applications or other purposes. Here, to provide a not-renally cleared therapeutics for renally impaired hyperuricemia patients, we designed a multi-tiered framework MatWISE that fuses state-of-the-art natural language processing, semantic relationship mapping, and machine learning to automate the complex process of material discovery from a sea of scientific literatures published until December of 2022, and successfully identified and optimized δ-MnO 2 into an orally administered, nonabsorbable uric acid (UA) lowering biomaterial. δ-MnO 2 had superior serum and urine UA-lowering effect in three hyperuricemia mouse models, by comparing with a standard of care drug. δ-MnO 2 is highly promising to serve as a safe and effective UA-lowering drug for renally impaired hyperuricemia patients. We demonstrated a new research paradigm for biomaterials that combining state-of-the-art machine learning techniques and a handful of experiments to discover a translationally relevant material from the massive existing research, for an unmet clinical need.
BibTeX
@misc{zeng2025automated,
  author = {Xiaodong Zeng and Jiahao Qiu and Xin Zhao and Kangfei Liu and Longfei Zhao and Junling An and Xiang Liang and Tianzhi Liu and Zhichao Wang and Weidi Xie and Mengdi Wang and Jie Luo and Shiyi Zhang},
  title = {{Automated Discovery of Therapeutic Biomaterial for Renally Impaired Hyperuricemia Patients by Natural Language Processing and Machine Learning}},
  year = {2025},
  doi = {10.1101/2025.03.05.641578},
  howpublished = {bioRxiv},
}
Figure