Published papers

Background: Traditional regression models struggle with non-linear data, limiting their ability to predict health outcomes such as activities of daily living (ADL) in stroke survivors. Machine learning (ML) methods offer improved accuracy but often lack interpretability. Explainable ML (XML) algorithms, including SHapley Additive exPlanations (SHAP), promise both precision and interpretability. This study aimed to develop interpretable XML-based prediction models for discharge ADL in subacute stroke survivors and to identify key prognostic factors. Methods: We retrospectively analyzed 980 subacute stroke survivors admitted to the Tokyo Bay Rehabilitation Hospital between March 2015 and September 2019. Thirty-three explanatory variables, including age, Functional Independence Measure (FIM) subitems, physical function measures, and nutritional assessments, were used to predict discharge FIM scores and FIM gain (motor, cognitive, total). Predictive models were developed using six algorithms (Elastic Net, Extra Trees, LightGBM, Random Forest, XGBoost, and their combinations - Mixed), and accuracy was evaluated using the coefficient of determination (R²) and root mean squared error (RMSE). Variable importance was quantified using SHAP scores. Results: Maximum R² values for motor, cognitive, and discharge total FIM scores were 0.776, 0.630, and 0.758, respectively. Those for FIM gain were 0.604, 0.413, and 0.536, respectively. SHAP scores highlighted the importance of age, trunk function, and grip strength. Furthermore, low-difficulty FIM motor subitems were significant predictors of total scores, whereas high-difficulty subitems were critical for FIM gain. Conclusions: XML achieved predictive performance comparable to reported parsimonious ML models, while providing higher clinical granular interpretability. By utilizing SHAP values, this approach overcomes the ‘black box’ nature of standard ML, providing transparent and actionable insights into how individual functional deficits contribute to outcomes. Notably, the predictive accuracy for FIM gain improved significantly compared to previous studies. Even without explicitly incorporating domain knowledge about the relationships between variables, XML was able to generate clinically relevant prediction models that accounted for these relationships. Thus, the examination of ADL prognostic factors using XML is clinically meaningful and has promising implications for real-world clinical applications.

Routine echocardiography at one month after diagnosis is the current standard for screening coronary artery abnormalities (CAA), a major complication of Kawasaki disease. The study aimed to develop and validate models to predict CAA and assess whether routine echocardiography could be safely reduced in low-risk patients. Two prospective multicenter Japanese registries were utilized: PEACOCK (development/internal validation) and Post-RAISE (external validation). Variables obtained within one week of diagnosis were used to predict CAA at one month after diagnosis, defined as a maximum coronary artery Z score (Zmax) ≥ 2. The models included simple models using the previous maximum Z score only, logistic regression models, and machine learning models (LightGBM and XGBoost). Discrimination, calibration, and clinical utility were assessed. Among 4,973 PEACOCK and 2,438 Post-RAISE patients, the CAA incidence was 5.5% and 6.8%, respectively. Twenty-two models were developed using 29 variables. For external validation, a simple model using the maximum Z score at week 1 produced an area under the curve (AUC) of 0.79; adding other variables or using more complex models did not increase the AUC by more than 0.02. The models failed to efficiently reduce the number of echocardiographic examinations while minimizing missed cased of CAA. Conclusion: Until superior predictors are identified, routine echocardiography at one month after diagnosis should remain the standard practice.

Modern clinical oncology faces an unprecedented data complexity that exceeds human analytical capacity, making artificial intelligence (AI) integration essential rather than optional. This review examines the dual impact of AI on productivity enhancement and creative discovery in cancer care. We trace the evolution from traditional machine learning to deep learning and transformer-based foundation models, analyzing their clinical applications. AI enhances productivity by automating diagnostic tasks, streamlining documentation, and accelerating research workflows across imaging modalities and clinical data processing. More importantly, AI enables creative discovery by integrating multimodal data to identify computational biomarkers, performing unsupervised phenotyping to reveal hidden patient subgroups, and accelerating drug development. Finally, we introduce the FUTURE-AI framework, outlining the essential requirements for translating AI models into clinical practice. This ensures the responsible deployment of AI, which augments rather than replaces clinical judgment, while maintaining patient-centered care.

Digital transformation of healthcare is rapidly progressing. Digital transformation improves the quality of services and access to health information for users, reduces the workload and associated costs for healthcare providers, and supports clinical decision-making. Data and artificial intelligence (AI) play a key role in this process. The AI used for this purpose is called medical AI. Medical AI is currently undergoing a shift from task-specific to general-purpose models. Large language models have the potential to systematize existing medical knowledge in a standardized way. The usage of AI in medicine is not limited to digital transformation; it plays a pivotal role in fundamentally changing the state of medical science. This approach, known as "AI for Medical Science," focuses on pioneering a form of medical science that predicts the onset and progression of disease based on the underlying causes of disease. The key to such predictive medicine is the concept of "states," which can be sought through machine learning. Using states instead of symptoms not only dramatically improves the accuracy of identification (diagnosis) and prediction (prognosis) but also potentially pioneers P4 medicine by integrating it with empirical knowledge and theories based on natural principles.

Melanoma cells display distinct intrinsic phenotypic states. Here, we seek to characterize the molecular regulation of these states using multi-omic analyses of whole exome, transcriptome, microRNA, long non-coding RNA and DNA methylation data together with reverse-phase protein array data on a panel of 68 highly annotated early passage melanoma cell lines. We demonstrate that clearly defined cancer cell intrinsic transcriptomic programs are maintained in melanoma cells ex vivo and remain highly conserved within melanoma tumors, are associated with distinct immune features within tumors, and differentially correlate with checkpoint inhibitor and adoptive T cell therapy efficacy. Through integrative analyses we demonstrate highly complex multi-omic regulation of melanoma cell intrinsic programs that provide key insights into the molecular maintenance of phenotypic states. These findings have implications for cancer biology and the identification of new therapeutic strategies. Further, these deeply characterized cell lines will serve as an invaluable resource for future research in the field.

Background: During the COVID-19 pandemic, there was an increasing need to expand diagnostic testing in hospitals. At Keio University Hospital (KUH), clinical staff were concerned that the demand for PCR testing might exceed the capacity of the Clinical Laboratory. In response, basic researchers at Keio University School of Medicine (KUSM) set out to build a new, collaborative, PCR testing system. To be authorized to perform such diagnostic PCR testing, KUSM registered its core laboratory as an external clinical laboratory (ECL). Methods: In the pandemic, there was a pressure to build the PCR system quickly. Speed required discussions that developed a shared understanding of the unprecedented, new KUH/KUSM PCR system. To design, construct, and archive the new PCR testing system, we used a systems engineering (SE) approach. This included diagram visualization of functional flows and application of the Unified Architecture Framework (UAF), both of which are often used in system building. We considered daily demand for PCR testing at KUH and KUSM, and daily COVID-19 infections in Japan. Results: We operated the collaborative PCR testing system from August 2020 to June 2022. Given public health insurance reimbursement policies, KUH focused on individuals with suspicious symptoms, while the ECL at KUSM screened samples from asymptomatic individuals. KUSM performed about half as many tests as KUH. Interviewing KUH staff revealed that diagrams helped promote a better understanding of the KUH/KUSM PCR testing system. Conclusion: When designing temporary systems that may be repurposed in the future, we suggest using an SE approach with diagrams and UAF perspectives. This approach will enable stakeholders to understand what is being proposed to be built, and facilitate achieving an informed consensus on the proposed system. We suggest that SE approaches should be widely used in projects that involve building and operating complex, collaborative systems, and documenting the process.

Melanoma represents the deadliest skin cancer. Recent therapeutic developments, including targeted and immune therapies have revolutionized clinical management and improved patient outcome. This progress was achieved by rigorous molecular and functional studies followed by robust clinical trials. The identification of key genomic alterations and gene expression profiles have propelled the understanding of distinct characteristics within melanoma subtypes. The aim of this review is to summarize and highlight the main genetic and epigenetic findings of melanomas and highlight their pathological and therapeutic importance.

From the beginning of the COVID-19 pandemic, the demand for diagnostic and screening tests has exceeded supply. Although the proportion of vaccinated people has increased in wealthier countries, breakthrough infections have occurred amid the emergence of new variants. Pooled-sample COVID-19 testing using saliva has been proposed as an efficient, inexpensive, and non-invasive method to allow larger-scale testing, especially in a screening setting. In this study, we aimed to evaluate pooled RT-qPCR saliva testing and to compare the results with individual tests. Employees of Philips Japan, Ltd. were recruited to participate in COVID-19 screening from October to December 2020. Asymptomatic individuals (n = 824) submitted self-collected saliva samples. Samples were tested for the presence of SARS-CoV-2 by RT-qPCR in both 10-sample pools and individual tests. We also surveyed participants regarding their thoughts and behaviors after the PCR screening project. Two of the 824 individuals were positive by RT-qPCR. In the pooled testing, one of these two had no measurable Ct value, but showed an amplification trend at the end of the PCR cycle. Both positive individuals developed cold-like symptoms, but neither required hospitalization. Of the 824 participants, 471 responded to our online questionnaire. Overall, while respondents agreed that PCR screening should be performed regularly, the majority were willing to undergo PCR testing only when it was provided for free or at low cost. In conclusion, pooled testing of saliva samples can support frequent large-scale screening that is rapid, efficient, and inexpensive.

The dynamic evolution of chromatin state patterns during metastasis, their relationship with bona fide genetic drivers, and their therapeutic vulnerabilities are not completely understood. Combinatorial chromatin state profiling of 46 melanoma samples reveals an association of NRAS mutants with bivalent histone H3 lysine 27 trimethylation (H3K27me3) and Polycomb repressive complex 2. Reprogramming of bivalent domains during metastasis occurs on master transcription factors of a mesenchymal phenotype, including ZEB1, TWIST1, and CDH1. Resolution of bivalency using pharmacological inhibition of EZH2 decreases invasive capacity of melanoma cells and markedly reduces tumor burden in vivo, specifically in NRAS mutants. Coincident with bivalent reprogramming, the increased expression of pro-metastatic and melanocyte-specific cell-identity genes is associated with exceptionally wide H3K4me3 domains, suggesting a role for this epigenetic element. Overall, we demonstrate that reprogramming of bivalent and broad domains represents key epigenetic alterations in metastatic melanoma and that EZH2 plus MEK inhibition may provide a promising therapeutic strategy for NRAS mutant melanoma patients.

The year 2020 will be remembered for the coronavirus disease 2019 (COVID-19) pandemic, which continues to affect the whole world. Early and accurate identification of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is fundamental to combat the disease. Among the current diagnostic tests, real-time reverse transcriptase–polymerase chain reaction (RT-qPCR) is the most reliable and frequently used method. Herein, we discuss the interpretation of RT-qPCR results relative to viral infectivity. Although nasopharyngeal swab samples are often used for RT-qPCR testing, they require collection by trained medical staff. Saliva samples are emerging as an inexpensive and efficient alternative for large-scale screening. Pooled-sample testing of saliva has been applied for mass screening of SARS-CoV-2 infection. Current policies recommend isolating people with borderline cycle threshold (Ct) values (35

Although treatment with immune checkpoint inhibitors provides promising benefit for patients with cancer, optimal use is encumbered by high resistance rates and requires a thorough understanding of resistance mechanisms. We observed that tumors treated with PD-1/PD-L1 blocking antibodies develop resistance through the upregulation of CD38, which is induced by all-trans retinoic acid and IFNβ in the tumor microenvironment. In vitro and in vivo studies demonstrate that CD38 inhibits CD8+ T-cell function via adenosine receptor signaling and that CD38 or adenosine receptor blockade are effective strategies to overcome the resistance. Large data sets of human tumors reveal expression of CD38 in a subset of tumors with high levels of basal or treatment-induced T-cell infiltration, where immune checkpoint therapies are thought to be most effective. These findings provide a novel mechanism of acquired resistance to immune checkpoint therapy and an opportunity to expand their efficacy in cancer treatment. SIGNIFICANCE: CD38 is a major mechanism of acquired resistance to PD-1/PD-L1 blockade, causing CD8+ T-cell suppression. Coinhibition of CD38 and PD-L1 improves antitumor immune response. Biomarker assessment in patient cohorts suggests that a combination strategy is applicable to a large percentage of patients in whom PD-1/PD-L1 blockade is currently indicated.

PURPOSE. Metastatic uveal melanoma (UM) has a very poor prognosis and no effective therapy. Despite remarkable advances in treatment of cutaneous melanoma, UM remains recalcitrant to chemotherapy, small-molecule kinase inhibitors, and immune-based therapy. METHODS. We assessed two sets of oxidative phosphorylation (OxPhos) genes within 9858 tumors across 31 cancer types. An OxPhos inhibitor was used to characterize differential metabolic programming of highly metastatic monosomy 3 (M3) UM. Seahorse analysis and global metabolomics profiling were done to identify metabolic vulnerabilities. Analyses of UM TCGA data set were performed to determine expressions of key OxPhos effectors in M3 and non-M3 UM. We used targeted knockdown of succinate dehydrogenase A (SDHA) to determine the role of SDHA in M3 UM in conferring resistance to OxPhos inhibition. RESULTS. We identified UM to have among the highest median OxPhos levels and showed that M3 UM exhibits a distinct metabolic profile. M3 UM shows markedly low succinate levels and has highly increased levels of SDHA, the enzyme that couples the tricarboxylic acid cycle with OxPhos by oxidizing (lowering) succinate. We showed that SDHA-high M3 UM have elevated expression of key OxPhos molecules, exhibit abundant mitochondrial reserve respiratory capacity, and are resistant to OxPhos antagonism, which can be reversed by SDHA knockdown. CONCLUSIONS. Our study has identified a critical metabolic program within poor prognostic M3 UM. In addition to the heightened mitochondrial functional capacity due to elevated SDHA, M3 UM SDHA-high mediate resistance to therapy that is reversible with targeted treatment.

The COVID-19 pandemic was unprecedented and forced countries to organise and combine resources and develop coordinated responses. Temporary systems were established with the support of researchers and other stakeholders. Dr Junna Oba, Assistant Professor at the Department of Extended Intelligence for Medicine at Keio University School of Medicine (KUSM) is devising methodologies for developing adjustable temporary systems that are quick and efficient to build, apply and dissemble when emergency situations have passed. KUSM helped to respond to the COVID-19 crisis by collaborating with basic science researches to reinforce the PCR testing system in hospitals. Among those involved in this effort were Professor Masako Toriya of Keio University Global Research Institute (KGRI) and Professor Koichi Matsuo from the KUSM's Collaborative Research Resources group. The researchers built from scratch a workflow for effective testing and information sharing. Since this time, the team has been working to build a cooperative network, developing a novel approach to planning, analysing and recording various processes and strategies to address emergency situations that will be shared with the wider healthcare community and policy makers. The researchers used systems engineering methods to launch an unprecedented collaborative system that offers the ability to gain an overall picture of the interrelationships between departments that are working together on the diagnosis, treatment and management of COVID-19.

Innate inflammatory features have been found in melanoma tumors from patients at all stages, and molecular analysis has identified definitive inflammatory proteins expressed by tumors cells in patients who presents the worst prognosis. We have previously observed weakened outcomes in patients with constitutive expression of inducible nitric oxide synthase (iNOS), macrophage migration inhibitory factor (MIF) and improved outcomes with CD74 expression in stage III melanoma. In our current study, we tested our hypothesis on CD74-regulated inflammatory markers’ expression in stage IV melanoma tumors whether the signature is associated with survival outcome and/or risk of developing CNS metastasis. We retrospectively identified 315 patients with stage IV melanoma. In a tissue microarray (TMA), we examined the expression of cells with CD74, its receptor MIF, and downstream inflammatory markers iNOS, nitrotyrosine (NT), cyclooxygenase (COX)-2 and microsomal prostaglandin E synthase-1 (mPGES1). We analyzed the association of those inflammatory markers with overall survival time (OS) and time to CNS metastasis using Kaplan–Meier survival analyses. Our data validates CD74 as a useful prognostic tumor cell protein marker associated with favorable OS as in stage III melanomas, while the tumor NT expression strongly predicts an increased risk of developing CNS metastasis (p = 0.0008) in those patients.

Sentinel lymph node biopsy (SLNB) is the standard of care for staging melanoma. However, limited research has been carried out on the prognostic value of SLNB in patients with primary cutaneous melanoma in Asian countries. The objective is to evaluate the efficacy of SLNB in Japanese patients with primary cutaneous melanoma and to elucidate whether reverse transcriptase (RT)-PCR analysis of sentinel lymph nodes (SLNs) is valuable for predicting patient outcome. A total of 101 patients with primary cutaneous melanoma underwent SLNB at the Department of Dermatology, Kyushu University (Fukuoka, Japan), between May 2001 and December 2009. The removed nodes were stained with hematoxylin-eosin and with immunohistochemical stains for HMB-45, tyrosinase, MART-1, and MITF and multiple-mRNA marker (MART-1, tyrosinase, and GP-100) RT-PCR assays were conducted. The following clinicopathological variables were evaluated: age, sex, histological type, tumor site, Breslow thickness, disease-free survival (DFS), and melanoma-specific survival (MSS). Several parameters were analyzed for DFS and MSS using the Kaplan-Meier method and Cox proportional hazards model. The success rate of identifying SLNs was 98% (99 of 101 cases). Tumor-positive SLNs were significantly correlated with higher Breslow thickness, stage, tumor subtype, and tumor site. Patients with tumor-positive SLNs had a significantly shorter MSS and DFS than those with tumor-negative SLNs (P=0.0153 and 0.0004, respectively). Patients with at least two positive markers in the RT-PCR assay had a significantly shorter DFS than those with less than one marker (P=0.013). SLNB and multimarker RT-PCR analysis are useful for predicting the prognosis of patients with melanoma.

This exploratory study was carried out to determine the expression levels of hepatocyte growth factor (HGF), insulin-like growth factor 1, thyroid-stimulating hormone (TSH), and leptin in serum and tumor samples from patients with uveal melanoma and to investigate the potential association of these expression levels with disease progression and patient survival. Seventeen patients, including nine nonmetastatic and eight metastatic, were included in the study. Eighteen healthy individuals served as controls. The levels of these four proteins in serum and tissue samples were determined by enzyme-linked immunosorbent assays and immunohistochemical staining, respectively. Associations between protein levels and survival, disease progression, and other clinicopathological factors were analyzed statistically. Serum levels of HGF were significantly higher and TSH levels were lower in uveal melanoma patients than in healthy individuals, but the level of neither protein differed significantly between metastatic and nonmetastatic groups. Of the four proteins tested, only serum TSH was significantly associated with patient survival. No correlation was observed between the tissue and serum levels of each protein. The levels of HGF in serum may be markers of uveal melanoma development. The prognostic and predictive values of these potential markers need to be determined in a larger cohort.

Melanomas of the choroid, ciliary body, and iris of the eye are collectively known as uveal melanomas. These cancers represent 5% of all melanoma diagnoses in the United States, and their age-adjusted risk is 5 per 1 million population. These less frequent melanomas are dissimilar to their more common cutaneous melanoma relative, with differing risk factors, primary treatment, anatomic spread, molecular changes, and responses to systemic therapy. Once uveal melanoma becomes metastatic, therapy options are limited and are often extrapolated from cutaneous melanoma therapies despite the routine exclusion of patients with uveal melanoma from clinical trials. Clinical trials directed at uveal melanoma have been completed or are in progress, and data from these well designed investigations will help guide future directions in this orphan disease. Cancer 2016;122:2299-2312. (c) 2016 American Cancer Society.