Shingo Tsuji
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Shingo Tsuji has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
- Tadatoshi Takayama
Department of Digestive Surgery, Nihon University School of Medicine, Tokyo 173-8610, Japan
4 shared publications - Mitsuhiko Moriyama
Division of Gastroenterology and Hepatology, Department of Medicine, Nihon University School of Medicine, 30-1 Oyaguchi-Kamicho, Itabashi-ku, Tokyo 173-8610, Japan
3 shared publications - Hayato Abe
Division of Digestive Surgery, Department of Surgery Nihon University School of Medicine Tokyo Japan
2 shared publications Hiroaki Kitano
The Systems Biology Institute, Tokyo, Japan
2 shared publicationsHiroshi Tanaka
Institute of Education, Tokyo Medical and Dental University, Tokyo, Japan
2 shared publicationsHiroyuki Aburatani
Research Center for Advanced Science and Technology, The University of Tokyo, Tokyo, Japan
2 shared publications
Latest Publications
- Open Access
- 25 September 2026
- Published by MDPI AG in Pharmaceuticals
- Vol. 19  (10) , 1522
Multilayered Prioritization of Graph Attention Autoencoder–Derived Drug Repurposing Candidates for COVID-19
Background/Objectives: The COVID-19 pandemic highlighted the critical need for drug repurposing. While computational graph-based approaches show promise, single analytical methods often fail to capture the complex pathological mechanisms needed for clinical translation. We aimed to prioritize COVID-19 repurposing candidates by combining computational prediction with orthogonal lines of evidence. Methods: We developed a Graph Attention Autoencoder (GATE) framework with multilayered evaluation integrating computational prediction, real-world data, and phenome analyses. Network-based embeddings identified candidates through latent-space similarity within biomedical knowledge graphs. Clinical relevance was assessed by disproportionality analysis (DPA) of an adverse event reporting system, and biological plausibility through gene expression and pathway profiling. Results: Our GATE framework identified 16 candidate drug entities, including 12 already used or recognized in the context of COVID-19 or its symptoms, consistent with the model recovering known drug–disease associations. The remaining four had minimal prior COVID-19 associations (cilastatin, megestrol, drotrecogin alfa, and ethacrynic acid). Of these four, the multilayered evaluation prioritized two: DPA showed significant inverse reporting associations for COVID-19-related adverse event terms for cilastatin (reporting odds ratio [ROR]: 0.19, 95% confidence interval [CI]: 0.09–0.43) and megestrol (ROR: 0.63, 95% CI: 0.42–0.94). Pathway profiling indicated that cilastatin shares molecular signatures with drugs investigated in COVID-19 clinical trials and megestrol with compounds reported as COVID-19 therapeutic candidates; gene expression analyses suggest cilastatin may exert antiviral and anti-inflammatory effects through DPEP1 inhibition. Conclusions: By integrating GATE-based prediction, real-world evidence, and phenome-level pathway profiling, this study enabled multidimensional candidate prioritization that single analytical methods do not provide. These findings, including the proposed mechanisms of action, are computational predictions that await experimental confirmation. They nonetheless provide mechanistic working hypotheses that may streamline subsequent preclinical studies. Prioritizing clinically approved drugs with established safety profiles may facilitate subsequent clinical evaluation, although the risks specific to each drug would need to be weighed. The framework offers a generalizable strategy for rapid candidate identification during pandemics and diseases with unmet therapeutic needs.
- 25 September 2026
- Published by MDPI AG in Pharmaceuticals
- Vol. 19  (10) , 1522
Open AccessMultilayered Prioritization of Graph Attention Autoencoder–Derived Drug Repurposing Candidates for COVID-19
Background/Objectives: The COVID-19 pandemic highlighted the critical need for drug repurposing. While computational graph-based approaches show promise, single analytical methods often fail to capture the complex pathological mechanisms needed for clinical translation. We aimed to prioritize COVID-19 repurposing candidates by combining computational prediction with orthogonal lines of evidence. Methods: We developed a Graph Attention Autoencoder (GATE) framework with multilayered evaluation integrating computational prediction, real-world data, and phenome analyses. Network-based embeddings identified candidates through latent-space similarity within biomedical knowledge graphs. Clinical relevance was assessed by disproportionality analysis (DPA) of an adverse event reporting system, and biological plausibility through gene expression and pathway profiling. Results: Our GATE framework identified 16 candidate drug entities, including 12 already used or recognized in the context of COVID-19 or its symptoms, consistent with the model recovering known drug–disease associations. The remaining four had minimal prior COVID-19 associations (cilastatin, megestrol, drotrecogin alfa, and ethacrynic acid). Of these four, the multilayered evaluation prioritized two: DPA showed significant inverse reporting associations for COVID-19-related adverse event terms for cilastatin (reporting odds ratio [ROR]: 0.19, 95% confidence interval [CI]: 0.09–0.43) and megestrol (ROR: 0.63, 95% CI: 0.42–0.94). Pathway profiling indicated that cilastatin shares molecular signatures with drugs investigated in COVID-19 clinical trials and megestrol with compounds reported as COVID-19 therapeutic candidates; gene expression analyses suggest cilastatin may exert antiviral and anti-inflammatory effects through DPEP1 inhibition. Conclusions: By integrating GATE-based prediction, real-world evidence, and phenome-level pathway profiling, this study enabled multidimensional candidate prioritization that single analytical methods do not provide. These findings, including the proposed mechanisms of action, are computational predictions that await experimental confirmation. They nonetheless provide mechanistic working hypotheses that may streamline subsequent preclinical studies. Prioritizing clinically approved drugs with established safety profiles may facilitate subsequent clinical evaluation, although the risks specific to each drug would need to be weighed. The framework offers a generalizable strategy for rapid candidate identification during pandemics and diseases with unmet therapeutic needs.
- Open Access
- 10 November 2024
- Published by Qeios Ltd
Review of: "AI-Generated Hallmarks of Aging and Cancer: A Computational Approach Using Causal Emergence and Dependency Networks"
The authors developed the AI-based hallmark generator and validated the gene set with computational approaches. Especially, causal emergence has attracted much attention in the field of complex sciences. It is very intriguing that the authors utilize...
- 10 November 2024
- Published by Qeios Ltd
Open AccessReview of: "AI-Generated Hallmarks of Aging and Cancer: A Computational Approach Using Causal Emergence and Dependency Networks"
The authors developed the AI-based hallmark generator and validated the gene set with computational approaches. Especially, causal emergence has attracted much attention in the field of complex sciences. It is very intriguing that the authors utilize...
- Open Access
- 3 May 2021
- Published by Springer Nature in Alzheimer's Research & Therapy
- Vol. 13  (1) , 1-15
Artificial intelligence-based computational framework for drug-target prioritization and inference of novel repositionable drugs for Alzheimer’s disease
Background: Identifying novel therapeutic targets is crucial for the successful development of drugs. However, the cost to experimentally identify therapeutic targets is huge and only approximately 400 genes are targets for FDA-approved drugs. As a result, it is inevitable to develop powerful computational tools that can identify potential novel therapeutic targets. Fortunately, the human protein-protein interaction network (PIN) could be a useful resource to achieve this objective. Methods: In this study, we developed a deep learning-based computational framework that extracts low-dimensional representations of high-dimensional PIN data. Our computational framework uses latent features and state-of-the-art machine learning techniques to infer potential drug target genes. Results: We applied our computational framework to prioritize novel putative target genes for Alzheimer’s disease and successfully identified key genes that may serve as novel therapeutic targets (e.g., DLG4, EGFR, RAC1, SYK, PTK2B, SOCS1). Furthermore, based on these putative targets, we could infer repositionable candidate-compounds for the disease (e.g., tamoxifen, bosutinib, and dasatinib). Conclusions: Our deep learning-based computational framework could be a powerful tool to efficiently prioritize new therapeutic targets and enhance the drug repositioning strategy.
- 3 May 2021
- Published by Springer Nature in Alzheimer's Research & Therapy
- Vol. 13  (1) , 1-15
Open AccessArtificial intelligence-based computational framework for drug-target prioritization and inference of novel repositionable drugs for Alzheimer’s disease
Background: Identifying novel therapeutic targets is crucial for the successful development of drugs. However, the cost to experimentally identify therapeutic targets is huge and only approximately 400 genes are targets for FDA-approved drugs. As a result, it is inevitable to develop powerful computational tools that can identify potential novel therapeutic targets. Fortunately, the human protein-protein interaction network (PIN) could be a useful resource to achieve this objective. Methods: In this study, we developed a deep learning-based computational framework that extracts low-dimensional representations of high-dimensional PIN data. Our computational framework uses latent features and state-of-the-art machine learning techniques to infer potential drug target genes. Results: We applied our computational framework to prioritize novel putative target genes for Alzheimer’s disease and successfully identified key genes that may serve as novel therapeutic targets (e.g., DLG4, EGFR, RAC1, SYK, PTK2B, SOCS1). Furthermore, based on these putative targets, we could infer repositionable candidate-compounds for the disease (e.g., tamoxifen, bosutinib, and dasatinib). Conclusions: Our deep learning-based computational framework could be a powerful tool to efficiently prioritize new therapeutic targets and enhance the drug repositioning strategy.
- Open Access
- 31 October 2020
- Published by International Research and Cooperation Association for Bio & Socio-Sciences Advancement (IRCA-BSSA) in BioScience Trends
- Vol. 14  (5) , 368-375
High platelet count as a poor prognostic factor for liver cancer patients without cirrhosis
A low platelet count, one of parameters of portal hypertension, is clinically a predictor of postoperative mortality, while platelets induce tumor development during growth factor secretion. In this study, we retrospectively investigated whether high platelet count negatively affects the survival of patients with hepatocellular carcinoma (HCC). Patients undergoing initial and curative resection for HCC were included. Surgical outcomes were compared between the high platelet (platelet count ≥ 20 × 104/μL) and control (< 20 × 104/μL) groups in patients without cirrhosis and between the low platelet (< 10 × 104/μL) and control (≥ 10 × 104/μL) groups in patients with cirrhosis. Among patients without cirrhosis, tumor was larger (P < 0.001) and tumor thrombus was more frequent (P < 0.001) in the high-platelet group than in the control group. After a median follow-up period of 3.1 years (range 0.2-16.2), median overall survival was 6.3 years (95% confidence interval [CI], 5.3-7.8) and 7.6 years (6.6-10.9) in the high-platelet (n = 273) and control (n = 562) groups, respectively (P = 0.027). Among patients with cirrhosis, liver function was worse (P < 0.001) and varices were more frequent (P < 0.001) in the low-platelet group. The median overall survival of patients in the low-platelet group (n = 172) was significantly shorter than that of patients in the control group (n = 275) (4.5 years [95% CI, 3.7–6.0] vs. 5.9 years [4.5-7.5], P = 0.038). Taken together, thrombocytopenia indicates poor prognosis in HCC patients with cirrhosis, while thrombocytosis is a poor prognostic predictor for those without cirrhosis.
- 31 October 2020
- Published by International Research and Cooperation Association for Bio & Socio-Sciences Advancement (IRCA-BSSA) in BioScience Trends
- Vol. 14  (5) , 368-375
Open AccessHigh platelet count as a poor prognostic factor for liver cancer patients without cirrhosis
A low platelet count, one of parameters of portal hypertension, is clinically a predictor of postoperative mortality, while platelets induce tumor development during growth factor secretion. In this study, we retrospectively investigated whether high platelet count negatively affects the survival of patients with hepatocellular carcinoma (HCC). Patients undergoing initial and curative resection for HCC were included. Surgical outcomes were compared between the high platelet (platelet count ≥ 20 × 104/μL) and control (< 20 × 104/μL) groups in patients without cirrhosis and between the low platelet (< 10 × 104/μL) and control (≥ 10 × 104/μL) groups in patients with cirrhosis. Among patients without cirrhosis, tumor was larger (P < 0.001) and tumor thrombus was more frequent (P < 0.001) in the high-platelet group than in the control group. After a median follow-up period of 3.1 years (range 0.2-16.2), median overall survival was 6.3 years (95% confidence interval [CI], 5.3-7.8) and 7.6 years (6.6-10.9) in the high-platelet (n = 273) and control (n = 562) groups, respectively (P = 0.027). Among patients with cirrhosis, liver function was worse (P < 0.001) and varices were more frequent (P < 0.001) in the low-platelet group. The median overall survival of patients in the low-platelet group (n = 172) was significantly shorter than that of patients in the control group (n = 275) (4.5 years [95% CI, 3.7–6.0] vs. 5.9 years [4.5-7.5], P = 0.038). Taken together, thrombocytopenia indicates poor prognosis in HCC patients with cirrhosis, while thrombocytosis is a poor prognostic predictor for those without cirrhosis.
- Open Access
Artificial intelligence based computational framework for drug-target prioritization and inference of novel repositionable drugs for Alzheimer’s disease
Background Identification of novel therapeutic targets is a key for successful drug development. However, the cost to experimentally identify therapeutic targets is huge and only 400 genes are targets for FDA-approved drugs. Therefore, it is inevitable to develop powerful computational tools to identify potential novel therapeutic targets. Because proteins make their functions together with their interacting partners, a protein-protein interaction network (PIN) in human could be a useful resource to build computational tools to investigate potential targets for therapeutic drugs. Network embedding methods, especially deep-learning based methods would be useful tools to extract an informative low-dimensional latent space that contains enough information required to fully represent original high-dimensional non-linear data of PINs. Results In this study, we developed a deep learning based computational framework that extracts low-dimensional latent space embedded in high-dimensional data of the human PIN and uses the features in the latent space (latent features) to infer potential novel targets for therapeutic drugs. We examined the relationships between the latent features and the representative network metrics and found that the network metrics can explain a large number of the latent features, while several latent features do not correlate with all the network metrics. The results indicate that the features are likely to capture information that the representative network metrics can not capture, while the latent features also can capture information obtained from the network metrics. Our computational framework uses the latent features together with state-of-the-art machine learning techniques to infer potential drug target genes. We applied our computational framework to prioritized novel putative target genes for Alzheimer’s disease and successfully identified key genes for potential novel therapeutic targets (e.g., DLG4, EGFR, RAC1, SYK, PTK2B, SOCS1). Furthermore, based on these putative targets, we inferred repositionable candidate-compounds for the disease (e.g., Tamoxifen, Bosutinib, and Dasatinib) Discussions Our computational framework could be powerful computational tools to efficiently prioritize new therapeutic targets and drug repositioning. It is pertinent to note here that our computational platform is easily applicable to investigate novel potential targets and repositionable compounds for any diseases, especially for rare diseases.
Open AccessArtificial intelligence based computational framework for drug-target prioritization and inference of novel repositionable drugs for Alzheimer’s disease
Background Identification of novel therapeutic targets is a key for successful drug development. However, the cost to experimentally identify therapeutic targets is huge and only 400 genes are targets for FDA-approved drugs. Therefore, it is inevitable to develop powerful computational tools to identify potential novel therapeutic targets. Because proteins make their functions together with their interacting partners, a protein-protein interaction network (PIN) in human could be a useful resource to build computational tools to investigate potential targets for therapeutic drugs. Network embedding methods, especially deep-learning based methods would be useful tools to extract an informative low-dimensional latent space that contains enough information required to fully represent original high-dimensional non-linear data of PINs. Results In this study, we developed a deep learning based computational framework that extracts low-dimensional latent space embedded in high-dimensional data of the human PIN and uses the features in the latent space (latent features) to infer potential novel targets for therapeutic drugs. We examined the relationships between the latent features and the representative network metrics and found that the network metrics can explain a large number of the latent features, while several latent features do not correlate with all the network metrics. The results indicate that the features are likely to capture information that the representative network metrics can not capture, while the latent features also can capture information obtained from the network metrics. Our computational framework uses the latent features together with state-of-the-art machine learning techniques to infer potential drug target genes. We applied our computational framework to prioritized novel putative target genes for Alzheimer’s disease and successfully identified key genes for potential novel therapeutic targets (e.g., DLG4, EGFR, RAC1, SYK, PTK2B, SOCS1). Furthermore, based on these putative targets, we inferred repositionable candidate-compounds for the disease (e.g., Tamoxifen, Bosutinib, and Dasatinib) Discussions Our computational framework could be powerful computational tools to efficiently prioritize new therapeutic targets and drug repositioning. It is pertinent to note here that our computational platform is easily applicable to investigate novel potential targets and repositionable compounds for any diseases, especially for rare diseases.
- Open Access
- 28 November 2017
- Published by Springer Nature in BMC Gastroenterology
- Vol. 17  (1) , 133
A prediction model for the grade of liver fibrosis using magnetic resonance elastography
Liver stiffness measurement (LSM) has recently become available for assessment of liver fibrosis. We aimed to develop a prediction model for liver fibrosis using clinical variables, including LSM. We performed a prospective study to compare liver fibrosis grade with fibrosis score. LSM was measured using magnetic resonance elastography in 184 patients that underwent liver resection, and liver fibrosis grade was diagnosed histologically after surgery. Using the prediction model established in the training group, we validated the classification accuracy in the independent test group. First, we determined a cut-off value for stratifying fibrosis grade using LSM in 122 patients in the training group, and correctly diagnosed fibrosis grades of 62 patients in the test group with a total accuracy of 69.3%. Next, on least absolute shrinkage and selection operator analysis in the training group, LSM (r = 0.687, P < 0.001), indocyanine green clearance rate at 15 min (ICGR15) (r = 0.527, P < 0.001), platelet count (r = –0.537, P < 0.001) were selected as variables for the liver fibrosis prediction model. This prediction model applied to the test group correctly diagnosed 32 of 36 (88.8%) Grade I (F0 and F1) patients, 13 of 18 (72.2%) Grade II (F2 and F3) patients, and 7 of 8 (87.5%) Grade III (F4) patients in the test group, with a total accuracy of 83.8%. The prediction model based on LSM, ICGR15, and platelet count can accurately and reproducibly predict liver fibrosis grade.
- 28 November 2017
- Published by Springer Nature in BMC Gastroenterology
- Vol. 17  (1) , 133
Open AccessA prediction model for the grade of liver fibrosis using magnetic resonance elastography
Liver stiffness measurement (LSM) has recently become available for assessment of liver fibrosis. We aimed to develop a prediction model for liver fibrosis using clinical variables, including LSM. We performed a prospective study to compare liver fibrosis grade with fibrosis score. LSM was measured using magnetic resonance elastography in 184 patients that underwent liver resection, and liver fibrosis grade was diagnosed histologically after surgery. Using the prediction model established in the training group, we validated the classification accuracy in the independent test group. First, we determined a cut-off value for stratifying fibrosis grade using LSM in 122 patients in the training group, and correctly diagnosed fibrosis grades of 62 patients in the test group with a total accuracy of 69.3%. Next, on least absolute shrinkage and selection operator analysis in the training group, LSM (r = 0.687, P < 0.001), indocyanine green clearance rate at 15 min (ICGR15) (r = 0.527, P < 0.001), platelet count (r = –0.537, P < 0.001) were selected as variables for the liver fibrosis prediction model. This prediction model applied to the test group correctly diagnosed 32 of 36 (88.8%) Grade I (F0 and F1) patients, 13 of 18 (72.2%) Grade II (F2 and F3) patients, and 7 of 8 (87.5%) Grade III (F4) patients in the test group, with a total accuracy of 83.8%. The prediction model based on LSM, ICGR15, and platelet count can accurately and reproducibly predict liver fibrosis grade.
- 1 February 2017
- Published by American Society of Clinical Oncology (ASCO) in Journal of Clinical Oncology
- Vol. 35  (4_suppl) , 307
Genetic difference between multicentric carcinogenesis and intrahepatic metastasis of multiple hepatocellular carcinoma.
307 Background: Multiple hepatocellular carcinoma (HCC) is categorized into 2 types; multicentric carcinogeneis (MC) and intrahepatic metastasis (IM). Although discrimination of the types of multiple HCC is of clinical importance, it is difficult to determine it correctly even after histological diagnosis. Methods: In order to classify multiple HCC into MC and IM, we compared pairs of multiple HCC and matched non-tumoral liver samples from 22 patients who were clinically diagnosed MC and 8 patients with IM with regard to molecular aberrations using next generation sequencer. Results: Exome sequence showed that among the 22 patients with clinical MC, 7 patients were genetically determined as having IM (gIM) because of the concordance of mutations, while all the patients with clinical IM were as having MC (gMC). In 15 patients with gIM, 125 mutations (90.6%) per patient were common, which were consistent with the result of a pair of primary HCC, and only a few passenger mutations were different between IM specimens suggesting that IM pairs developed from the common ancestor. On the other hand, 138 somatic mutations per tumor, and no common somatic mutations were identified in the 15 patients with gMC. Clinically, vascular invasion was significantly frequent in gIM group ( P = 0.007). On the other hand, after a median follow-up of 2.56 years (range; 0.27−6.32 years), the median overall survivals were 3.7 years (95%CI, 1.26–6.14) and 3.37 years (2.02−6.32, P = 0.712) in the patients with IM and MC, respectively. The 3-year rates of overall survival were 71.7% and 65.1% in the two groups, respectively. Conclusions: Taken together, genetic comparison of mutations in a pair of HCCs makes it possible to classify MC and IM, which is available to make treatment plan for multiple HCC patients.
- 1 February 2017
- Published by American Society of Clinical Oncology (ASCO) in Journal of Clinical Oncology
- Vol. 35  (4_suppl) , 307
Genetic difference between multicentric carcinogenesis and intrahepatic metastasis of multiple hepatocellular carcinoma.
307 Background: Multiple hepatocellular carcinoma (HCC) is categorized into 2 types; multicentric carcinogeneis (MC) and intrahepatic metastasis (IM). Although discrimination of the types of multiple HCC is of clinical importance, it is difficult to determine it correctly even after histological diagnosis. Methods: In order to classify multiple HCC into MC and IM, we compared pairs of multiple HCC and matched non-tumoral liver samples from 22 patients who were clinically diagnosed MC and 8 patients with IM with regard to molecular aberrations using next generation sequencer. Results: Exome sequence showed that among the 22 patients with clinical MC, 7 patients were genetically determined as having IM (gIM) because of the concordance of mutations, while all the patients with clinical IM were as having MC (gMC). In 15 patients with gIM, 125 mutations (90.6%) per patient were common, which were consistent with the result of a pair of primary HCC, and only a few passenger mutations were different between IM specimens suggesting that IM pairs developed from the common ancestor. On the other hand, 138 somatic mutations per tumor, and no common somatic mutations were identified in the 15 patients with gMC. Clinically, vascular invasion was significantly frequent in gIM group ( P = 0.007). On the other hand, after a median follow-up of 2.56 years (range; 0.27−6.32 years), the median overall survivals were 3.7 years (95%CI, 1.26–6.14) and 3.37 years (2.02−6.32, P = 0.712) in the patients with IM and MC, respectively. The 3-year rates of overall survival were 71.7% and 65.1% in the two groups, respectively. Conclusions: Taken together, genetic comparison of mutations in a pair of HCCs makes it possible to classify MC and IM, which is available to make treatment plan for multiple HCC patients.
- Open Access
- 15 September 2012
- Published by Springer Nature in BMC Systems Biology
- Vol. 6  (1) , 124
A simple knowledge-based mining method for exploring hidden key molecules in a human biomolecular network
Background: In the functional genomics analysis domain, various methodologies are available for interpreting the results produced by high-throughput biological experiments. These methods commonly use a list of genes as an analysis input, and most of them produce a more complicated list of genes or pathways as the results of the analysis. Although there are several network-based methods, which detect key nodes in the network, the results tend to include well-studied, major hub genes. Results: To mine the molecules that have biological meaning but to fewer degrees than major hubs, we propose, in this study, a new network-based method for selecting these hidden key molecules based on virtual information flows circulating among the input list of genes. The human biomolecular network was constructed from the Pathway Commons database, and a calculation method based on betweenness centrality was newly developed. We validated the method with the ErbB pathway and applied it to practical cancer research data. We were able to confirm that the output genes, despite having fewer edges than major hubs, have biological meanings that were able to be invoked by the input list of genes. Conclusions: The developed method, named NetHiKe (Network-based Hidden Key molecule miner), was able to detect potential key molecules by utilizing the human biomolecular network as a knowledge base. Thus, it is hoped that this method will enhance the progress of biological data analysis in the whole-genome research era.
- 15 September 2012
- Published by Springer Nature in BMC Systems Biology
- Vol. 6  (1) , 124
Open AccessA simple knowledge-based mining method for exploring hidden key molecules in a human biomolecular network
Background: In the functional genomics analysis domain, various methodologies are available for interpreting the results produced by high-throughput biological experiments. These methods commonly use a list of genes as an analysis input, and most of them produce a more complicated list of genes or pathways as the results of the analysis. Although there are several network-based methods, which detect key nodes in the network, the results tend to include well-studied, major hub genes. Results: To mine the molecules that have biological meaning but to fewer degrees than major hubs, we propose, in this study, a new network-based method for selecting these hidden key molecules based on virtual information flows circulating among the input list of genes. The human biomolecular network was constructed from the Pathway Commons database, and a calculation method based on betweenness centrality was newly developed. We validated the method with the ErbB pathway and applied it to practical cancer research data. We were able to confirm that the output genes, despite having fewer edges than major hubs, have biological meanings that were able to be invoked by the input list of genes. Conclusions: The developed method, named NetHiKe (Network-based Hidden Key molecule miner), was able to detect potential key molecules by utilizing the human biomolecular network as a knowledge base. Thus, it is hoped that this method will enhance the progress of biological data analysis in the whole-genome research era.