Hongji Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Hongji Yang has a total of 184 co-authors. Here are the top co-authors by number of shared publications:
- Chen Li
Aalborg University
6 shared publications Charlotte Poile
University of Leicester
5 shared publications- Dean Anthony Fennell
University of Leicester
5 shared publications - Edward J. Hollox
University of Leicester
5 shared publications - Joanna Dzialo
UK National Institute for Health Research Biomedical Research Centre & Cancer Research UK Experimental Cancer Medicine Centre, Mesothelioma Research Programme, University of Leicester, Leicester, UK
5 shared publications Peter Wells-Jordan
University of Leicester
5 shared publications
Latest Publications
- Open Access
The Impact of AI on Game Development: A Scoping Review of Transformations, Career Implications, and Emerging Challenges
Artificial intelligence (AI) is increasingly transforming the game industry by reshaping development processes, production workflows, and career structures. In this study, we present a scoping literature review examining the impact of AI on game development and employment within the industry, with particular attention to the perspectives of displacement, task transformation, and productivity effects. Drawing on academic research and industry reports published between 2020 and 2026, while retaining earlier foundational literature where relevant, this study analyzes how AI technologies such as generative AI, machine learning, procedural content generation, and large language models are integrated across major stages of game production, including game art, production management, programming, and live operations. Our findings indicate that AI can support efficiency, accelerate development cycles, and enable new forms of personalization and creative experimentation. At the same time, AI adoption may alter workforce dynamics by increasing the automation potential of routine and entry-level tasks, especially in areas such as asset creation and quality assurance. Rather than fully replacing workers, however, AI more commonly reconfigures job roles and increases demand for hybrid skill sets that combine technical knowledge, creativity, and AI literacy. Our study also identifies emerging opportunities in AI-assisted design, data analytics, and machine learning integration. Despite these benefits, AI adoption introduces important challenges related to employment polarization, intellectual property, workforce adaptation, ethical governance, and market concentration. We conclude that AI is fundamentally reshaping the game industry and highlight the need for balanced strategies that support innovation while preserving creative diversity, sustainable career pathways, and equitable access to technological opportunities. Overall, the strongest evidence concerns AI capabilities, workflow changes, and task-level transformation, whereas direct evidence of actual employment outcomes and long-term occupational change remains limited and uneven.
Open AccessThe Impact of AI on Game Development: A Scoping Review of Transformations, Career Implications, and Emerging Challenges
Artificial intelligence (AI) is increasingly transforming the game industry by reshaping development processes, production workflows, and career structures. In this study, we present a scoping literature review examining the impact of AI on game development and employment within the industry, with particular attention to the perspectives of displacement, task transformation, and productivity effects. Drawing on academic research and industry reports published between 2020 and 2026, while retaining earlier foundational literature where relevant, this study analyzes how AI technologies such as generative AI, machine learning, procedural content generation, and large language models are integrated across major stages of game production, including game art, production management, programming, and live operations. Our findings indicate that AI can support efficiency, accelerate development cycles, and enable new forms of personalization and creative experimentation. At the same time, AI adoption may alter workforce dynamics by increasing the automation potential of routine and entry-level tasks, especially in areas such as asset creation and quality assurance. Rather than fully replacing workers, however, AI more commonly reconfigures job roles and increases demand for hybrid skill sets that combine technical knowledge, creativity, and AI literacy. Our study also identifies emerging opportunities in AI-assisted design, data analytics, and machine learning integration. Despite these benefits, AI adoption introduces important challenges related to employment polarization, intellectual property, workforce adaptation, ethical governance, and market concentration. We conclude that AI is fundamentally reshaping the game industry and highlight the need for balanced strategies that support innovation while preserving creative diversity, sustainable career pathways, and equitable access to technological opportunities. Overall, the strongest evidence concerns AI capabilities, workflow changes, and task-level transformation, whereas direct evidence of actual employment outcomes and long-term occupational change remains limited and uneven.
- Open Access
- 7 July 2026
- Published by Elsevier in Journal of Systems and Software
- Vol. 242
An exploratory study on LLM-generated code and comments in code repositories
- 7 July 2026
- Published by Elsevier in Journal of Systems and Software
- Vol. 242
Open AccessAn exploratory study on LLM-generated code and comments in code repositories
- Open Access
- 5 June 2026
- Published by Springer Nature in Research Square
Genomic Entropy confers an aggressive Phenotype in Mesothelioma
Exposure to asbestos is the primary cause of mesothelioma, however the processes underpinning its initiation and evolutionary trajectory remain elusive. To address this knowledge gap, we conducted whole exome sequencing and phylogenetic analysis of 277 mesotheliomas. Using multidimensional Dirichlet process clustering we inferred phylogenies, uncovering mutation order dominated by early 13q,1p, 8p, 6q and 22q somatic copy number alterations. Mesotheliomas exhibit a spectrum of genomic intra-tumor heterogeneity spanning linear (L) to highly branched (B) tree topologies, quantifiable by the Weighted Shannon Entropy (WSE). Somatic copy number alterations emerge earlier than single nucleotide variants (SNVs), and at a much faster relative rate in L-mesotheliomas. BAP1 is the only positively selected driver in L-mesotheliomas, whereas, B-mesotheliomas harbor multiple positively selected drivers and an aggressive phenotype, including a bias of loss of RPL5, and MTAP deletion linked to breakpoint enriched with long interspersed nuclear elements. L-trees had lower WSE, but greater T-cell inflammation and tertiary lymphoid structures, despite having a lower neoantigen burden. WSE is higher in larger volume, immunologically cold mesotheliomas, and is associated with shorter survival. In summary, progressive genomic heterogeneity is a defining feature of more aggressive mesothelioma yet illuminates opportunities for precision therapeutic intervention.
- 5 June 2026
- Published by Springer Nature in Research Square
Open AccessGenomic Entropy confers an aggressive Phenotype in Mesothelioma
Exposure to asbestos is the primary cause of mesothelioma, however the processes underpinning its initiation and evolutionary trajectory remain elusive. To address this knowledge gap, we conducted whole exome sequencing and phylogenetic analysis of 277 mesotheliomas. Using multidimensional Dirichlet process clustering we inferred phylogenies, uncovering mutation order dominated by early 13q,1p, 8p, 6q and 22q somatic copy number alterations. Mesotheliomas exhibit a spectrum of genomic intra-tumor heterogeneity spanning linear (L) to highly branched (B) tree topologies, quantifiable by the Weighted Shannon Entropy (WSE). Somatic copy number alterations emerge earlier than single nucleotide variants (SNVs), and at a much faster relative rate in L-mesotheliomas. BAP1 is the only positively selected driver in L-mesotheliomas, whereas, B-mesotheliomas harbor multiple positively selected drivers and an aggressive phenotype, including a bias of loss of RPL5, and MTAP deletion linked to breakpoint enriched with long interspersed nuclear elements. L-trees had lower WSE, but greater T-cell inflammation and tertiary lymphoid structures, despite having a lower neoantigen burden. WSE is higher in larger volume, immunologically cold mesotheliomas, and is associated with shorter survival. In summary, progressive genomic heterogeneity is a defining feature of more aggressive mesothelioma yet illuminates opportunities for precision therapeutic intervention.
- 6 April 2026
- Published by Springer Nature in Cluster Computing
- Vol. 29 (4) , 249
TATEA: a time-aware and topology-aware framework for entity alignment in temporal knowledge graphs
- 6 April 2026
- Published by Springer Nature in Cluster Computing
- Vol. 29 (4) , 249
TATEA: a time-aware and topology-aware framework for entity alignment in temporal knowledge graphs
- Open Access
- 21 July 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 1-13
Constitutive inflammation and epithelial-mesenchymal transition dictate sensitivity to nivolumab in CONFIRM: a placebo-controlled, randomised phase III trial
Leveraging adaptive tumour immunity to control mesothelioma via immune checkpoint blockade is now a standard therapeutic approach. However, the determinants of sensitivity remain elusive. Low non-synonymous mutation burden and programmed death-ligand 1 expression, an abundance of immunosuppressive immune cell infiltration, and 9p21 deletion should all mitigate responses to therapy. To address this knowledge gap, we conducted a double blind, placebo-controlled, randomized phase III trial of the PD1 inhibitor, nivolumab (ClinicalTrial.gov registration: NCT03063450). After 37.2 months of follow-up, the primary endpoint of progression free-survival, but not overall survival was met. The nivolumab response rate was 10.3%, and related grade 3 or above adverse events occurred in 20.4% versus 7.2% for placebo. Progression-free and overall survival were longer in nivolumab-treated responders versus non-responders. In an exploratory multiomic analysis, blinded whole exome, transcriptome and multiplex immune profiling were used to interrogate R- versus NR-subgroups. Non-synonymous and neoantigen mutation burden were no different between groups, however R-mesotheliomas were infiltrated with activated CD8+ T- and CD19+ B-lymphocytes, organised into tertiary lymphoid structures. B-cell infiltration correlated with pro-inflammatory chemokines including IL24 and CCL19. Conversely, epithelial-mesenchymal transition and mitosis were associated with resistance to nivolumab. These findings illuminate features which can be leveraged to advance precision immunotherapy in this rare cancer setting.
- 21 July 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 1-13
Open AccessConstitutive inflammation and epithelial-mesenchymal transition dictate sensitivity to nivolumab in CONFIRM: a placebo-controlled, randomised phase III trial
Leveraging adaptive tumour immunity to control mesothelioma via immune checkpoint blockade is now a standard therapeutic approach. However, the determinants of sensitivity remain elusive. Low non-synonymous mutation burden and programmed death-ligand 1 expression, an abundance of immunosuppressive immune cell infiltration, and 9p21 deletion should all mitigate responses to therapy. To address this knowledge gap, we conducted a double blind, placebo-controlled, randomized phase III trial of the PD1 inhibitor, nivolumab (ClinicalTrial.gov registration: NCT03063450). After 37.2 months of follow-up, the primary endpoint of progression free-survival, but not overall survival was met. The nivolumab response rate was 10.3%, and related grade 3 or above adverse events occurred in 20.4% versus 7.2% for placebo. Progression-free and overall survival were longer in nivolumab-treated responders versus non-responders. In an exploratory multiomic analysis, blinded whole exome, transcriptome and multiplex immune profiling were used to interrogate R- versus NR-subgroups. Non-synonymous and neoantigen mutation burden were no different between groups, however R-mesotheliomas were infiltrated with activated CD8+ T- and CD19+ B-lymphocytes, organised into tertiary lymphoid structures. B-cell infiltration correlated with pro-inflammatory chemokines including IL24 and CCL19. Conversely, epithelial-mesenchymal transition and mitosis were associated with resistance to nivolumab. These findings illuminate features which can be leveraged to advance precision immunotherapy in this rare cancer setting.
- 9 April 2025
- Published by Springer Nature
- p. 3-12
A Creative Computing Approach to Dimentionalising the DIKW Model: Bridging the Gap Between Concepts and Practice
- 9 April 2025
- Published by Springer Nature
- p. 3-12
A Creative Computing Approach to Dimentionalising the DIKW Model: Bridging the Gap Between Concepts and Practice
- Open Access
- 31 January 2025
- Published by Journal of Internet Technology in 網際網路技術學刊
- Vol. 26 (1) , 77-87
Using Big Data from Internet to Improve Land Use Transition Effects on Eco-Environment
The intricate relationship between land use and ecological changes is a critical area of eco-environmental research. However, there is a notable lack of quantitative, analytical studies that explore this relationship in depth. This study aims to bridge this gap by applying a contrastive learning approach to a practical land use application, utilizing big data. We analyzed 2,197 publications from the Web of Science Core Collection spanning from 1998 to 2022. Our methodology correlates textual and visual data, enhancing the categorization and thematic analysis of existing literature. The contrastive learning technique uncovers nuanced relationships between textual descriptions and visual data, providing a comprehensive view of research trends. Key findings include an increasing volume of publications, indicating growing interest and potential in this field; predominant focus areas such as the relationship between climate change and land use, dynamic land use modelling, urbanization impacts, land use management, ecosystem services, and biodiversity. Emerging research hotspots identified through contrastive learning include the interactions between urban thermal environments and land use changes, land use effects on biological communities and soil organic matter, and the relationship between ecological conditions and land use in vulnerable regions. This study not only maps the current research landscape but also employs a contrastive learning approach to predict future research directions, thus enriching the methodological framework for analyzing land use applications with big data.
- 31 January 2025
- Published by Journal of Internet Technology in 網際網路技術學刊
- Vol. 26 (1) , 77-87
Open AccessUsing Big Data from Internet to Improve Land Use Transition Effects on Eco-Environment
The intricate relationship between land use and ecological changes is a critical area of eco-environmental research. However, there is a notable lack of quantitative, analytical studies that explore this relationship in depth. This study aims to bridge this gap by applying a contrastive learning approach to a practical land use application, utilizing big data. We analyzed 2,197 publications from the Web of Science Core Collection spanning from 1998 to 2022. Our methodology correlates textual and visual data, enhancing the categorization and thematic analysis of existing literature. The contrastive learning technique uncovers nuanced relationships between textual descriptions and visual data, providing a comprehensive view of research trends. Key findings include an increasing volume of publications, indicating growing interest and potential in this field; predominant focus areas such as the relationship between climate change and land use, dynamic land use modelling, urbanization impacts, land use management, ecosystem services, and biodiversity. Emerging research hotspots identified through contrastive learning include the interactions between urban thermal environments and land use changes, land use effects on biological communities and soil organic matter, and the relationship between ecological conditions and land use in vulnerable regions. This study not only maps the current research landscape but also employs a contrastive learning approach to predict future research directions, thus enriching the methodological framework for analyzing land use applications with big data.
- 21 October 2024
- Published by Springer Nature in Cluster Computing
- Vol. 28 (1) , 1-13
A scalable rule learning approach fusing multiple sources of information
- 21 October 2024
- Published by Springer Nature in Cluster Computing
- Vol. 28 (1) , 1-13
A scalable rule learning approach fusing multiple sources of information
- Open Access
- 21 August 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-14
A gut microbiota rheostat forecasts responsiveness to PD-L1 and VEGF blockade in mesothelioma
Malignant mesothelioma is a rare tumour caused by asbestos exposure that originates mainly from the pleural lining or the peritoneum. Treatment options are limited, and the prognosis is dismal. Although immune checkpoint blockade (ICB) can improve survival outcomes, the determinants of responsiveness remain elusive. Here, we report the outcomes of a multi-centre phase II clinical trial (MiST4, NCT03654833) evaluating atezolizumab and bevacizumab (AtzBev) in patients with relapsed mesothelioma. We also use tumour tissue and gut microbiome sequencing, as well as tumour spatial immunophenotyping to identify factors associated with treatment response. MIST4 met its primary endpoint with 50% 12-week disease control, and the treatment was tolerable. Aneuploidy, notably uniparental disomy (UPD), homologous recombination deficiency (HRD), epithelial-mesenchymal transition and inflammation with CD68+ monocytes were identified as tumour-intrinsic resistance factors. The log-ratio of gut-resident microbial genera positively correlated with radiological response to AtzBev and CD8+ T cell infiltration, but was inversely correlated with UPD, HRD and tumour infiltration by CD68+ monocytes. In summary, a model is proposed in which both intrinsic and extrinsic determinants in mesothelioma cooperate to modify the tumour microenvironment and confer clinical sensitivity to AtzBev. Gut microbiota represent a potentially modifiable factor with potential to improve immunotherapy outcomes for individuals with this cancer of unmet need.
- 21 August 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-14
Open AccessA gut microbiota rheostat forecasts responsiveness to PD-L1 and VEGF blockade in mesothelioma
Malignant mesothelioma is a rare tumour caused by asbestos exposure that originates mainly from the pleural lining or the peritoneum. Treatment options are limited, and the prognosis is dismal. Although immune checkpoint blockade (ICB) can improve survival outcomes, the determinants of responsiveness remain elusive. Here, we report the outcomes of a multi-centre phase II clinical trial (MiST4, NCT03654833) evaluating atezolizumab and bevacizumab (AtzBev) in patients with relapsed mesothelioma. We also use tumour tissue and gut microbiome sequencing, as well as tumour spatial immunophenotyping to identify factors associated with treatment response. MIST4 met its primary endpoint with 50% 12-week disease control, and the treatment was tolerable. Aneuploidy, notably uniparental disomy (UPD), homologous recombination deficiency (HRD), epithelial-mesenchymal transition and inflammation with CD68+ monocytes were identified as tumour-intrinsic resistance factors. The log-ratio of gut-resident microbial genera positively correlated with radiological response to AtzBev and CD8+ T cell infiltration, but was inversely correlated with UPD, HRD and tumour infiltration by CD68+ monocytes. In summary, a model is proposed in which both intrinsic and extrinsic determinants in mesothelioma cooperate to modify the tumour microenvironment and confer clinical sensitivity to AtzBev. Gut microbiota represent a potentially modifiable factor with potential to improve immunotherapy outcomes for individuals with this cancer of unmet need.
- Open Access
- 25 May 2024
- Published by MDPI AG in Future Internet
- Vol. 16 (6) , 189
Dynamic Spatial–Temporal Self-Attention Network for Traffic Flow Prediction
Traffic flow prediction is considered to be one of the fundamental technologies in intelligent transportation systems (ITSs) with a tremendous application prospect. Unlike traditional time series analysis tasks, the key challenge in traffic flow prediction lies in effectively modelling the highly complex and dynamic spatiotemporal dependencies within the traffic data. In recent years, researchers have proposed various methods to enhance the accuracy of traffic flow prediction, but certain issues still persist. For instance, some methods rely on specific static assumptions, failing to adequately simulate the dynamic changes in the data, thus limiting their modelling capacity. On the other hand, some approaches inadequately capture the spatiotemporal dependencies, resulting in the omission of crucial information and leading to unsatisfactory prediction outcomes. To address these challenges, this paper proposes a model called the Dynamic Spatial–Temporal Self-Attention Network (DSTSAN). Firstly, this research enhances the interaction between different dimension features in the traffic data through a feature augmentation module, thereby improving the model’s representational capacity. Subsequently, the current investigation introduces two masking matrices: one captures local spatial dependencies and the other captures global spatial dependencies, based on the spatial self-attention module. Finally, the methodology employs a temporal self-attention module to capture and integrate the dynamic temporal dependencies of traffic data. We designed experiments using historical data from the previous hour to predict traffic flow conditions in the hour ahead, and the experiments were extensively compared to the DSTSAN model, with 11 baseline methods using four real-world datasets. The results demonstrate the effectiveness and superiority of the proposed approach.
- 25 May 2024
- Published by MDPI AG in Future Internet
- Vol. 16 (6) , 189
Open AccessDynamic Spatial–Temporal Self-Attention Network for Traffic Flow Prediction
Traffic flow prediction is considered to be one of the fundamental technologies in intelligent transportation systems (ITSs) with a tremendous application prospect. Unlike traditional time series analysis tasks, the key challenge in traffic flow prediction lies in effectively modelling the highly complex and dynamic spatiotemporal dependencies within the traffic data. In recent years, researchers have proposed various methods to enhance the accuracy of traffic flow prediction, but certain issues still persist. For instance, some methods rely on specific static assumptions, failing to adequately simulate the dynamic changes in the data, thus limiting their modelling capacity. On the other hand, some approaches inadequately capture the spatiotemporal dependencies, resulting in the omission of crucial information and leading to unsatisfactory prediction outcomes. To address these challenges, this paper proposes a model called the Dynamic Spatial–Temporal Self-Attention Network (DSTSAN). Firstly, this research enhances the interaction between different dimension features in the traffic data through a feature augmentation module, thereby improving the model’s representational capacity. Subsequently, the current investigation introduces two masking matrices: one captures local spatial dependencies and the other captures global spatial dependencies, based on the spatial self-attention module. Finally, the methodology employs a temporal self-attention module to capture and integrate the dynamic temporal dependencies of traffic data. We designed experiments using historical data from the previous hour to predict traffic flow conditions in the hour ahead, and the experiments were extensively compared to the DSTSAN model, with 11 baseline methods using four real-world datasets. The results demonstrate the effectiveness and superiority of the proposed approach.