Ion Petre
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
Ion Petre has a total of 189 co-authors. Here are the top co-authors by number of shared publications:
- Tero Harju
Department of Mathematics and Statistics, University of Turku, Finland
32 shared publications - Grzegorz Rozenberg
Universiteit Leiden
22 shared publications - Andrzej Ehrenfeucht
Department of Computer Science, University of Colorado at Boulder, Boulder, CO 80309-0347, USA
20 shared publications - David M. Prescott
University of Colorado
16 shared publications Grzegorz Rozenberg
11 shared publications- Sepinoud Azimi
Delft University of Technology
11 shared publications
Latest Publications
- Open Access
- 31 October 2026
- Published by Elsevier in Critical Reviews in Oncology/Hematology
- Vol. 228
Raman spectroscopy for glioma diagnosis toward precision neuro-oncology: A systematic review
- 31 October 2026
- Published by Elsevier in Critical Reviews in Oncology/Hematology
- Vol. 228
Open AccessRaman spectroscopy for glioma diagnosis toward precision neuro-oncology: A systematic review
- Open Access
- 26 October 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1) , 34019
Deep learning mitigates batch effects in Raman spectroscopy
Raman spectroscopy is widely used for its ability to provide rich molecular-level information from biological samples. However, the high sensitivity of Raman spectra can also unintentionally encode extraneous information, such as patient identity, instrument-specific signatures, environmental conditions, and sample handling protocols. This hidden structure, referred to as the batch effect, poses increasing challenges in machine learning applications, where it can lead to model overfitting, poor generalization, and even privacy risks when datasets are shared. In this work, we investigate the presence and impact of the batch effect in a Raman spectroscopy dataset derived from glioma tumor samples. We show that deep learning can accurately predict patient identity from individual spectra, revealing strong bias patterns in the data. To mitigate this issue, we explore several approaches and introduce an algorithmic framework based on adversarial deep learning. Our method effectively reduces the influence of confounding sample-specific signals while preserving information relevant to tumor classification. These findings highlight the need for careful bias control in Raman-based AI pipelines and provide tools to support more robust applications in biomedical spectroscopy.
- 26 October 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1) , 34019
Open AccessDeep learning mitigates batch effects in Raman spectroscopy
Raman spectroscopy is widely used for its ability to provide rich molecular-level information from biological samples. However, the high sensitivity of Raman spectra can also unintentionally encode extraneous information, such as patient identity, instrument-specific signatures, environmental conditions, and sample handling protocols. This hidden structure, referred to as the batch effect, poses increasing challenges in machine learning applications, where it can lead to model overfitting, poor generalization, and even privacy risks when datasets are shared. In this work, we investigate the presence and impact of the batch effect in a Raman spectroscopy dataset derived from glioma tumor samples. We show that deep learning can accurately predict patient identity from individual spectra, revealing strong bias patterns in the data. To mitigate this issue, we explore several approaches and introduce an algorithmic framework based on adversarial deep learning. Our method effectively reduces the influence of confounding sample-specific signals while preserving information relevant to tumor classification. These findings highlight the need for careful bias control in Raman-based AI pipelines and provide tools to support more robust applications in biomedical spectroscopy.
- Open Access
- 22 September 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (9) , e1014785
In-silico personalized protein–protein interaction networks prioritize candidate compounds for glioblastoma
The incidence of cancer continues to rise globally, with many tumor types remaining difficult to treat effectively. While existing drugs can alleviate symptoms or slow disease progression, their efficacy varies considerably between patients. To explore this challenge, we present a computational approach for identifying patient-specific candidate compounds. Our approach was developed for glioblastoma, but it can be applied to other types of cancer. Our study involves identifying possible disease-relevant proteins specific to each patient that underlie the generated networks, together with drug targets and proteins associated with cancer dependency. These networks are analyzed using centrality measures and Louvain community detection method. Then, further investigating the individual networks, network controllability analysis is applied to identify key regulatory targets within the network. These targets are filtered and ranked to prioritize candidate drugs for individualized treatment. By comparing results across patients and against networks based on generic data, we show substantial differences between patient-specific and generic network representations. Additionally, we track changes in patient-specific networks over time in five glioblastoma cases, exploring how molecular differences between primary and recurrent tumors may influence network structure and computational drug prioritization.
- 22 September 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (9) , e1014785
Open AccessIn-silico personalized protein–protein interaction networks prioritize candidate compounds for glioblastoma
The incidence of cancer continues to rise globally, with many tumor types remaining difficult to treat effectively. While existing drugs can alleviate symptoms or slow disease progression, their efficacy varies considerably between patients. To explore this challenge, we present a computational approach for identifying patient-specific candidate compounds. Our approach was developed for glioblastoma, but it can be applied to other types of cancer. Our study involves identifying possible disease-relevant proteins specific to each patient that underlie the generated networks, together with drug targets and proteins associated with cancer dependency. These networks are analyzed using centrality measures and Louvain community detection method. Then, further investigating the individual networks, network controllability analysis is applied to identify key regulatory targets within the network. These targets are filtered and ranked to prioritize candidate drugs for individualized treatment. By comparing results across patients and against networks based on generic data, we show substantial differences between patient-specific and generic network representations. Additionally, we track changes in patient-specific networks over time in five glioblastoma cases, exploring how molecular differences between primary and recurrent tumors may influence network structure and computational drug prioritization.
- Open Access
- 3 September 2026
- Published by Springer Nature in Natural Computing
- Vol. 25 (2) , 30
Multiset reaction systems with saturation and permanence
We propose an extension to multisets of the model of reaction systems, classically based on sets. In the proposed model, four main variations are present with respect to the classical model: (1) reactions (here called m -reactions) have multisets rather than sets in their reactant, inhibitor, and product components; (2) states are multisets of entities, rather than sets; (3) based on the numerosity of the entities in the state, several instances of non-conflicting m -reactions can be simultaneously applied, forming a block ; and (4) entities not consumed in the application of a block remain in the state, to which the cumulatively produced multiset of entities is added. Threshold phenomena (reaching a critical mass enabling new m -reactions or saturating the state with entities inhibiting m -reactions to be enabled) can therefore be observed, so that some context-independent behaviours can be observed, which are not possible in the classical case. The paper explores some formal consequences of the proposed model and illustrates its relevance by presenting a model of the dynamics of the heat shock response cellular regulatory mechanism.
- 3 September 2026
- Published by Springer Nature in Natural Computing
- Vol. 25 (2) , 30
Open AccessMultiset reaction systems with saturation and permanence
We propose an extension to multisets of the model of reaction systems, classically based on sets. In the proposed model, four main variations are present with respect to the classical model: (1) reactions (here called m -reactions) have multisets rather than sets in their reactant, inhibitor, and product components; (2) states are multisets of entities, rather than sets; (3) based on the numerosity of the entities in the state, several instances of non-conflicting m -reactions can be simultaneously applied, forming a block ; and (4) entities not consumed in the application of a block remain in the state, to which the cumulatively produced multiset of entities is added. Threshold phenomena (reaching a critical mass enabling new m -reactions or saturating the state with entities inhibiting m -reactions to be enabled) can therefore be observed, so that some context-independent behaviours can be observed, which are not possible in the classical case. The paper explores some formal consequences of the proposed model and illustrates its relevance by presenting a model of the dynamics of the heat shock response cellular regulatory mechanism.
- Open Access
Network-based analysis of glioblastoma identifies patient communities and cluster-specific biomarkers
Glioblastoma is an aggressive and highly heterogeneous brain tumor with poor prognosis despite multimodal treatment strategies. Understanding the molecular diversity of the disease is essential for improving tumor stratification and identifying potential therapeutic targets. In this study, we investigate whether network-based analysis can reveal biologically meaningful subgroups of glioblastoma tumors. Using RNA sequencing and mutation data from the TCGA-GBM cohort, we constructed patient-specific protein–protein interaction networks based on genes that are differentially expressed or harbor somatic mutations. These networks capture the molecular alterations associated with individual tumors within the context of the human interactome. We then derived similarities between tumors using a binary representation of network nodes and the Jaccard similarity metric, enabling the construction of a patient similarity graph. Community detection algorithms (Louvain and Leiden) were applied to this graph to identify clusters of tumors with similar molecular network profiles. Our analysis revealed six tumor communities characterized by distinct gene compositions and enriched biological processes. For each community, we identified candidate biomarkers and network hubs that may represent potential therapeutic targets. Several of the identified genes correspond to known drug targets, while others represent potential candidates for further investigation. These results illustrate how integrating molecular alterations with network-based modeling can help stratify glioblastoma tumors and uncover molecular mechanisms that may guide the development of more personalized therapeutic strategies.
Open AccessNetwork-based analysis of glioblastoma identifies patient communities and cluster-specific biomarkers
Glioblastoma is an aggressive and highly heterogeneous brain tumor with poor prognosis despite multimodal treatment strategies. Understanding the molecular diversity of the disease is essential for improving tumor stratification and identifying potential therapeutic targets. In this study, we investigate whether network-based analysis can reveal biologically meaningful subgroups of glioblastoma tumors. Using RNA sequencing and mutation data from the TCGA-GBM cohort, we constructed patient-specific protein–protein interaction networks based on genes that are differentially expressed or harbor somatic mutations. These networks capture the molecular alterations associated with individual tumors within the context of the human interactome. We then derived similarities between tumors using a binary representation of network nodes and the Jaccard similarity metric, enabling the construction of a patient similarity graph. Community detection algorithms (Louvain and Leiden) were applied to this graph to identify clusters of tumors with similar molecular network profiles. Our analysis revealed six tumor communities characterized by distinct gene compositions and enriched biological processes. For each community, we identified candidate biomarkers and network hubs that may represent potential therapeutic targets. Several of the identified genes correspond to known drug targets, while others represent potential candidates for further investigation. These results illustrate how integrating molecular alterations with network-based modeling can help stratify glioblastoma tumors and uncover molecular mechanisms that may guide the development of more personalized therapeutic strategies.
- Open Access
- 7 April 2026
- Published by Springer Nature in Diagnostic and Prognostic Research
- Vol. 10 (1) , 9
Raman spectroscopy as a diagnostic tool for glioma: protocol for a systematic review
Glioma is the most common primary brain tumor, demanding prompt and accurate diagnosis to guide therapy decisions. Conventional diagnostic methods such as
- 7 April 2026
- Published by Springer Nature in Diagnostic and Prognostic Research
- Vol. 10 (1) , 9
Open AccessRaman spectroscopy as a diagnostic tool for glioma: protocol for a systematic review
Glioma is the most common primary brain tumor, demanding prompt and accurate diagnosis to guide therapy decisions. Conventional diagnostic methods such as
- Open Access
- 5 November 2025
- Published by Springer Nature in Natural Computing
- Vol. 24 (4) , 1101-1117
Model checking for distributed reaction systems with temporal-epistemic properties
Reaction systems are a model of computation inspired by the biochemistry exhibited by living cells. This paper introduces the notion of agency as an extension to the reaction systems formalism, leading to distributed reaction systems. Adding agents in the reaction systems setting, allows for the natural modelling and representation of multi-agent and distributed systems. To support the specification of temporal-epistemic properties of distributed reaction systems, we introduce the logic rs ctlk and present experimental results of its associated model checking procedure run on a biological benchmark of within-cell signal transduction networks. The experimental results are encouraging despite the complexity of the rs ctlk model checking problem that is shown to be pspace -complete.
- 5 November 2025
- Published by Springer Nature in Natural Computing
- Vol. 24 (4) , 1101-1117
Open AccessModel checking for distributed reaction systems with temporal-epistemic properties
Reaction systems are a model of computation inspired by the biochemistry exhibited by living cells. This paper introduces the notion of agency as an extension to the reaction systems formalism, leading to distributed reaction systems. Adding agents in the reaction systems setting, allows for the natural modelling and representation of multi-agent and distributed systems. To support the specification of temporal-epistemic properties of distributed reaction systems, we introduce the logic rs ctlk and present experimental results of its associated model checking procedure run on a biological benchmark of within-cell signal transduction networks. The experimental results are encouraging despite the complexity of the rs ctlk model checking problem that is shown to be pspace -complete.
- Open Access
- 29 October 2025
- Published by Springer Nature in Natural Computing
- Vol. 24 (4) , 1011-1012
Preface
- 29 October 2025
- Published by Springer Nature in Natural Computing
- Vol. 24 (4) , 1011-1012
Open AccessPreface
- 29 October 2025
- Published by Springer Nature in Natural Computing
- Vol. 24 (4) , 917-918
Preface
- Open Access
- 4 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 119709-119725
Simulation and Analysis of Distributed Reaction Systems
Reaction systems (RSs) are a computational framework inspired by the interplay between biochemical interactions. Similarly to regulatory networks, reaction products can dynamically activate or inhibit other reactions. This provides a framework for discrete-time, interactive computation, where each state is determined by those reactions enabled in the immediately preceding state and by additional environmental interventions, if any. Since their introduction, RSs have been extended to study many aspects of complex systems: multi-agent collaborations, cause-effect relationships, model checking, and many others. The gap in the literature addressed in this paper is the lack of software tools for simulating and analysing distributed reaction systems (DRSs). We introduce a process algebraic approach to describing, simulating, and analysing DRSs. This allows designing complex models by re-using modular components in a well-structured and compositional way, and analysing them with the BioReSolve modelling software. We demonstrate our approach by experimenting with distributed Lotka-Volterra models, where multiple agents evolve according to their own periodic dynamics, but can also synchronise their cycles through diverse communication topologies.
- 4 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 119709-119725
Open AccessSimulation and Analysis of Distributed Reaction Systems
Reaction systems (RSs) are a computational framework inspired by the interplay between biochemical interactions. Similarly to regulatory networks, reaction products can dynamically activate or inhibit other reactions. This provides a framework for discrete-time, interactive computation, where each state is determined by those reactions enabled in the immediately preceding state and by additional environmental interventions, if any. Since their introduction, RSs have been extended to study many aspects of complex systems: multi-agent collaborations, cause-effect relationships, model checking, and many others. The gap in the literature addressed in this paper is the lack of software tools for simulating and analysing distributed reaction systems (DRSs). We introduce a process algebraic approach to describing, simulating, and analysing DRSs. This allows designing complex models by re-using modular components in a well-structured and compositional way, and analysing them with the BioReSolve modelling software. We demonstrate our approach by experimenting with distributed Lotka-Volterra models, where multiple agents evolve according to their own periodic dynamics, but can also synchronise their cycles through diverse communication topologies.