Yasushi Okuno

Aya Nakamura, Ryosuke Kojima, Yuji Okamoto et al.

Many diseases, including cancer and chronic conditions, require extended treatment periods and long-term strategies. Machine learning and AI research focusing on electronic health records (EHRs) have emerged to address this need. Effective treatment strategies involve more than capturing sequential changes in patient test values. It requires an explainable and clinically interpretable model by capturing the patient's internal state over time. In this study, we propose the "deep state-space analysis framework," using time-series unsupervised learning of EHRs with a deep state-space model. This framework enables learning, visualizing, and clustering of temporal changes in patient latent states related to disease progression. We evaluated our framework using time-series laboratory data from 12,695 cancer patients. By estimating latent states, we successfully discover latent states related to prognosis. By visualization and cluster analysis, the temporal transition of patient status and test items during state transitions characteristic of each anticancer drug were identified. Our framework surpasses existing methods in capturing interpretable latent space. It can be expected to enhance our comprehension of disease progression from EHRs, aiding treatment adjustments and prognostic determinations.

Kazuki Nakamura, Eiichiro Uchino, Noriaki Sato et al.

Early disease detection and prevention methods based on effective interventions are gaining attention. Machine learning technology has enabled precise disease prediction by capturing individual differences in multivariate data. Progress in precision medicine has revealed that substantial heterogeneity exists in health data at the individual level and that complex health factors are involved in the development of chronic diseases. However, it remains a challenge to identify individual physiological state changes in cross-disease onset processes because of the complex relationships among multiple biomarkers. Here, we present the health-disease phase diagram (HDPD), which represents a personal health state by visualizing the boundary values of multiple biomarkers that fluctuate early in the disease progression process. In HDPDs, future onset predictions are represented by perturbing multiple biomarker values while accounting for dependencies among variables. We constructed HDPDs for 11 non-communicable diseases (NCDs) from a longitudinal health checkup cohort of 3,238 individuals, comprising 3,215 measurement items and genetic data. Improvement of biomarker values to the non-onset region in HDPD significantly prevented future disease onset in 7 out of 11 NCDs. Our results demonstrate that HDPDs can represent individual physiological states in the onset process and be used as intervention goals for disease prevention.

Yusuke Watanabe, Yohei Kobashi, Takeshi Kojima et al.

Clinical decisions are often required under incomplete information. Clinical experts must identify whether available information is sufficient for judgment, as both premature conclusion and unnecessary abstention can compromise patient safety. To evaluate this capability of large language models (LLMs), we developed ClinDet-Bench, a benchmark based on clinical scoring systems that decomposes incomplete-information scenarios into determinable and undeterminable conditions. Identifying determinability requires considering all hypotheses about missing information, including unlikely ones, and verifying whether the conclusion holds across them. We find that recent LLMs fail to identify determinability under incomplete information, producing both premature judgments and excessive abstention, despite correctly explaining the underlying scoring knowledge and performing well under complete information. These findings suggest that existing benchmarks are insufficient to evaluate the safety of LLMs in clinical settings. ClinDet-Bench provides a framework for evaluating determinability recognition, leading to appropriate abstention, with potential applicability to medicine and other high-stakes domains, and is publicly available.

Atsuko Takagi, Mayumi Kamada, Eri Hamatani et al.

Drug repositioning holds great promise because it can reduce the time and cost of new drug development. While drug repositioning can omit various R&D processes, confirming pharmacological effects on biomolecules is essential for application to new diseases. Biomedical explainability in a drug repositioning model can support appropriate insights in subsequent in-depth studies. However, the validity of the XAI methodology is still under debate, and the effectiveness of XAI in drug repositioning prediction applications remains unclear. In this study, we propose GraphIX, an explainable drug repositioning framework using biological networks, and quantitatively evaluate its explainability. GraphIX first learns the network weights and node features using a graph neural network from known drug indication and knowledge graph that consists of three types of nodes (but not given node type information): disease, drug, and protein. Analysis of the post-learning features showed that node types that were not known to the model beforehand are distinguished through the learning process based on the graph structure. From the learned weights and features, GraphIX then predicts the disease-drug association and calculates the contribution values of the nodes located in the neighborhood of the predicted disease and drug. We hypothesized that the neighboring protein node to which the model gave a high contribution is important in understanding the actual pharmacological effects. Quantitative evaluation of the validity of protein nodes' contribution using a real-world database showed that the high contribution proteins shown by GraphIX are reasonable as a mechanism of drug action. GraphIX is a framework for evidence-based drug discovery that can present to users new disease-drug associations and identify the protein important for understanding its pharmacological effects from a large and complex knowledge base.

Nanae Aratake, Taisei Tosaki, Yuji Okamoto et al.

Clinical risk prediction using longitudinal medical data supports individualized care. Self-supervised foundation models have emerged as a promising approach for leveraging large-scale unlabeled healthcare records. In natural language processing, scaling laws suggest that larger models achieve predictably lower pretraining losses, supporting the foundation model paradigm. However, for structured medical data, characterized by a limited vocabulary and sparse observations, whether increasing model size consistently improves downstream predictions is unclear, as most studies evaluate only a single model scale. In this study, we evaluated the relationship between model scale and downstream task performance for structured medical foundation models. Using a random sample (2.3 million patients, 32 hospitals) from a nationwide 519-hospital Japanese claims database, we pretrained encoder-only Transformers at five scales (2.2M-101M parameters) for disease incidence and medication prediction. Downstream performance saturated at task-dependent thresholds: disease prediction benefited from larger models (32M-101M), whereas medication prediction saturated at 11M, reducing pretraining time by 178 h. Across all tasks, the best-performing model consistently outperformed a Light Gradient Boosting Machine baseline in the area under the precision-recall curve. These findings indicate that, unlike the monotonically decreasing pretraining loss, the optimal model size varied depending on task characteristics. This task-dependent saturation provides practical guidance for balancing predictive performance and computational cost in structured medical foundation models.

Kazuma Inoue, Ryosuke Kojima, Mayumi Kamada et al.

Biological data may be separated into primary data, such as gene expression, and secondary data, such as pathways and protein-protein interactions. Methods using secondary data to enhance the analysis of primary data are promising, because secondary data have background information that is not included in primary data. In this study, we proposed an end-to-end framework to integrally handle secondary data to construct a classification model for primary data. We applied this framework to cancer prognosis prediction using gene expression data and a biological network. Cross-validation results indicated that our model achieved higher accuracy compared with a deep neural network model without background biological network information. Experiments conducted in patient groups by cancer type showed improvement in ROC-area under the curve for many groups. Visualizations of high accuracy cancer types identified contributing genes and pathways by enrichment analysis. Known biomarkers and novel biomarker candidates were identified through these experiments.

Seungeon Lee, Takuto Koyama, Itsuki Maeda et al.

Therapeutic peptides have emerged as a pivotal modality in modern drug discovery, occupying a chemically and topologically rich space. While accurate prediction of their physicochemical properties is essential for accelerating peptide development, existing molecular language models rely on representations that fail to capture this complexity. Atom-level SMILES notation generates long token sequences and obscures cyclic topology, whereas amino-acid-level representations cannot encode the diverse chemical modifications central to modern peptide design. To bridge this representational gap, the Hierarchical Editing Language for Macromolecules (HELM) offers a unified framework enabling precise description of both monomer composition and connectivity, making it a promising foundation for peptide language modeling. Here, we propose HELM-BERT, the first encoder-based peptide language model trained on HELM notation. Based on DeBERTa, HELM-BERT is specifically designed to capture hierarchical dependencies within HELM sequences. The model is pre-trained on a curated corpus of 39,079 chemically diverse peptides spanning linear and cyclic structures. HELM-BERT significantly outperforms state-of-the-art SMILES-based language models in downstream tasks, including cyclic peptide membrane permeability prediction and peptide-protein interaction prediction. These results demonstrate that HELM's explicit monomer- and topology-aware representations offer substantial data-efficiency advantages for modeling therapeutic peptides, bridging a long-standing gap between small-molecule and protein language models.

Sho Oshima, Yuji Okamoto, Taisei Tosaki et al.

Graph Contrastive Learning (GCL) is a powerful self-supervised learning framework that performs data augmentation through graph perturbations, with growing applications in the analysis of biological networks such as Gene Regulatory Networks (GRNs). The artificial perturbations commonly used in GCL, such as node dropping, induce structural changes that can diverge from biological reality. This concern has contributed to a broader trend in graph representation learning toward augmentation-free methods, which view such structural changes as problematic and to be avoided. However, this trend overlooks the fundamental insight that structural changes from biologically meaningful perturbations are not a problem to be avoided but a rich source of information, thereby ignoring the valuable opportunity to leverage data from real biological experiments. Motivated by this insight, we propose SupGCL (Supervised Graph Contrastive Learning), a new GCL method for GRNs that directly incorporates biological perturbations from gene knockdown experiments as supervision. SupGCL is a probabilistic formulation that continuously generalizes conventional GCL, linking artificial augmentations with real perturbations measured in knockdown experiments and using the latter as explicit supervisory signals. To assess effectiveness, we train GRN representations with SupGCL and evaluate their performance on downstream tasks. The evaluation includes both node-level tasks, such as gene function classification, and graph-level tasks on patient-specific GRNs, such as patient survival hazard prediction. Across 13 tasks built from GRN datasets derived from patients with three cancer types, SupGCL consistently outperforms state-of-the-art baselines.

Kazuki Nakamura, Ryosuke Kojima, Eiichiro Uchino et al.

Clinical decision making regarding treatments based on personal characteristics leads to effective health improvements. Machine learning (ML) has been the primary concern of diagnosis support according to comprehensive patient information. However, the remaining prominent issue is the development of objective treatment processes in clinical situations. This study proposes a novel framework to plan treatment processes in a data-driven manner. A key point of the framework is the evaluation of the "actionability" for personal health improvements by using a surrogate Bayesian model in addition to a high-performance nonlinear ML model. We first evaluated the framework from the viewpoint of its methodology using a synthetic dataset. Subsequently, the framework was applied to an actual health checkup dataset comprising data from 3,132 participants, to improve systolic blood pressure values at the individual level. We confirmed that the computed treatment processes are actionable and consistent with clinical knowledge for lowering blood pressure. These results demonstrate that our framework could contribute toward decision making in the medical field, providing clinicians with deeper insights.