Justin Seby
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Justin Seby has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Alireza Bakhshi Zadi Mahmoodi
University of Oulu, Oulu, Pohjois-Pohjanmaa, Finland
1 shared publication - Arun Viswanathan
Cancer Research Program, BRIC-Rajiv Gandhi Center for Biotechnology (BRIC-RGCB), Thiruvananthapuram, Kerala State, 695014, India
1 shared publication - Danial Khosh Kholgh
Center for Ubiquitous Computing, University of Oulu, Oulu, Finland
1 shared publication Jaakko Suutala
1 shared publication- Jarkko Okkonen
Geological Survey of Finland, P.O. Box 97, FI-67101 Kokkola, Finland
1 shared publication - Kuzhuvelil Bhaskarannair Harikumar
Cancer Research Program, BRIC-Rajiv Gandhi Center for Biotechnology (BRIC-RGCB), Thiruvananthapuram, Kerala State, 695014, India
1 shared publication
Latest Publications
- Open Access
- 29 September 2026
- Published by Elsevier in SSRN Electronic Journal
Beyond Expensive Simulations: Trustworthy Machine Learning Surrogates for Scalable Geophysical Predictions
This study investigates the development and evaluation of statistical machine learning surrogates to accelerate computationally expensive numerical simulations in geophysics. We focus on creating surrogates capable of predicting soil temperature profiles at various depths. To assess model reliability, we explore predictive uncertainty using Gaussian Processes and Monte Carlo Dropout with multilayer perceptrons. We implement active learning using two complementary strategies: one driven by model uncertainty and the other by acquisition functions. These methods aim to strategically select the most informative training points, thereby reducing labeling effort while improving model performance. To address scalability, we apply both GPs and MLPs for spatial interpolation and extrapolation. By comparing surrogate model performance to the original simulator in terms of accuracy and computational efficiency, this research provides insights into improving the speed and reliability of geophysical simulations.
- 29 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessBeyond Expensive Simulations: Trustworthy Machine Learning Surrogates for Scalable Geophysical Predictions
This study investigates the development and evaluation of statistical machine learning surrogates to accelerate computationally expensive numerical simulations in geophysics. We focus on creating surrogates capable of predicting soil temperature profiles at various depths. To assess model reliability, we explore predictive uncertainty using Gaussian Processes and Monte Carlo Dropout with multilayer perceptrons. We implement active learning using two complementary strategies: one driven by model uncertainty and the other by acquisition functions. These methods aim to strategically select the most informative training points, thereby reducing labeling effort while improving model performance. To address scalability, we apply both GPs and MLPs for spatial interpolation and extrapolation. By comparing surrogate model performance to the original simulator in terms of accuracy and computational efficiency, this research provides insights into improving the speed and reliability of geophysical simulations.
- Open Access
Overweight status drives early tumor microenvironment reprogramming in pancreatic ductal adenocarcinoma: a cell-type-resolved Bayesian hierarchical modeling and interactome analysis
Obesity significantly increases the risk of pancreatic ductal adenocarcinoma, yet a comprehensive understanding of obesity-driven tumor microenvironment remodeling remains incomplete. We developed a cell-type-resolved transcriptomic model using bulk RNA sequencing data from the Clinical Proteomic Tumor Analysis Consortium cohort (n=140) stratified by body mass index. A custom functional gene signature database covering 65 immune and stromal cell types was manually constructed and validated through large-language model-assisted review. Cell-type-specific expression profiles were derived using BayesPrism deconvolution with matched single-cell RNA sequencing references, enabling high-resolution quantification of signature activity within each cell type. Body mass index-associated signatures were identified using a machine learning framework. Bulk pathway analysis showed deterioration of extracellular matrix homeostasis and primary immunodeficiency pathways with rising body mass index. Bayesian hierarchical modeling revealed cell-type-specific, non-linear dynamics: stromal populations in overweight individuals underwent coordinated extracellular matrix remodeling and inflammatory signaling that stabilized rather than intensified in obesity, while CD8-positive T cells showed dose-dependent activation progressing to chronic exhaustion. Spatial analysis showed that stromal-trapped CD8-positive T cells were positioned progressively closer to the cytokeratin-positive tumor boundary with rising body mass index. These findings establish overweight status as a critical tipping point in tumor microenvironment reprogramming, suggesting that early metabolic interventions may prevent irreversible microenvironmental deterioration. Abstract Figure:
Open AccessOverweight status drives early tumor microenvironment reprogramming in pancreatic ductal adenocarcinoma: a cell-type-resolved Bayesian hierarchical modeling and interactome analysis
Obesity significantly increases the risk of pancreatic ductal adenocarcinoma, yet a comprehensive understanding of obesity-driven tumor microenvironment remodeling remains incomplete. We developed a cell-type-resolved transcriptomic model using bulk RNA sequencing data from the Clinical Proteomic Tumor Analysis Consortium cohort (n=140) stratified by body mass index. A custom functional gene signature database covering 65 immune and stromal cell types was manually constructed and validated through large-language model-assisted review. Cell-type-specific expression profiles were derived using BayesPrism deconvolution with matched single-cell RNA sequencing references, enabling high-resolution quantification of signature activity within each cell type. Body mass index-associated signatures were identified using a machine learning framework. Bulk pathway analysis showed deterioration of extracellular matrix homeostasis and primary immunodeficiency pathways with rising body mass index. Bayesian hierarchical modeling revealed cell-type-specific, non-linear dynamics: stromal populations in overweight individuals underwent coordinated extracellular matrix remodeling and inflammatory signaling that stabilized rather than intensified in obesity, while CD8-positive T cells showed dose-dependent activation progressing to chronic exhaustion. Spatial analysis showed that stromal-trapped CD8-positive T cells were positioned progressively closer to the cytokeratin-positive tumor boundary with rising body mass index. These findings establish overweight status as a critical tipping point in tumor microenvironment reprogramming, suggesting that early metabolic interventions may prevent irreversible microenvironmental deterioration. Abstract Figure:
- Top Cited PapersOpen Access
- 8 January 2025
- Published by Elsevier in Decision Analytics Journal
- Vol. 14
A cognitive platform for collecting cyber threat intelligence and real-time detection using cloud computing
- 8 January 2025
- Published by Elsevier in Decision Analytics Journal
- Vol. 14
Top Cited PapersOpen AccessA cognitive platform for collecting cyber threat intelligence and real-time detection using cloud computing