Back to search

Institutes outside Greece

Computational omics translation models for designing better experimental disease models

Abstract

dc:description

Developing effective therapies relies heavily on pre-clinical models that closely mimic human diseases. Yet, today's models, whether animal-based or advanced cell cultures, often fail to translate reliably into meaningful clinical outcomes. This is largely due to key biological differences and an inability to fully capture disease complexity. This thesis addresses these challenges by creating computational tools that utilize large-scale biological (omics) data combined with machine learning to improve how we design and use these experimental systems. Throughout my work, I focus specifically on metabolic dysfunction-associated fatty liver disease (MAFLD) as a practical example.In Chapter 2, I designed a computational approach that pinpoints which important aspects of human biology are represented—or missing—in liver microphysiological systems (MPS). After comparing gene expression data from MAFLD patients with existing MPS models, it became clear that these systems do capture much of the disease-relevant human variability. However, they notably missed key molecular responses related to interferon signaling. By highlighting interferon as a missing piece, the analysis suggested that including interferon treatments in the MPS could greatly boost its clinical relevance. Additionally, through an unbiased computational optimization, I identified specific conditions—especially involving JAK-STAT signaling and hypoxia—that could further align MPS models with the diversity seen in actual patients.Chapter 3 focused on creating deep-learning tools capable of simulating detailed cell signaling processes and predicting how cells respond to drug treatments. Using neural networks based on biological interactomes, I successfully predicted transcription factor activities and even identified unexpected off-target effects of drugs. Through a detailed investigation of hepatocytes, I discovered critical interactions between TGFβ and interferon signaling in MAFLD. These insights confirmed earlier predictions and showed how deep learning can suggest and support new therapeutic hypotheses that would otherwise remain hidden.Finally, in Chapter 4, I tackled the challenge of translating biological insights between species and systems. Here, I developed AutoTransOP, a deep-learning framework capable of moving seamlessly between different biological contexts without relying on direct gene-to-gene mappings. This made it possible to translate findings across species or even entirely different cell systems. For example, AutoTransOP effectively translated data between human and animal models in lung fibrosis research and accurately predicted immune responses in HIV studies. When applied to the MAFLD data, this method reinforced the therapeutic importance of TGFβ signaling, making it possible to "humanize" results derived from experimental models.Together, this thesis provides practical computational methods that systematically refine and enhance the way experimental models of human disease are developed and interpreted. By integrating these computational techniques into preclinical research, the ultimate goal is to accelerate the discovery of new treatments that are more likely to succeed in clinical trials.

Degree

thesis:*
Grantor dc:publisher
Institutes outside Greece
Year dc:date
2025

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Meimetis, Nikolaos
  • Μεϊμέτης, Νικόλαος

Subjects

dc:subject × 28

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/59831

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Meimetis, Nikolaos; Μεϊμέτης, Νικόλαος. Computational omics translation models for designing better experimental disease models. Institutes outside Greece, 2025. http://hdl.handle.net/10442/hedi/59831