KTU PharmAI

Mission

We advance pharmacology by designing innovative AI and data-driven methods that enhance and accelerate traditional experimental workflows—from hypothesis generation to mechanism-of-action (MoA) inference and preclinical decision-making.

Who we are

PharmAI (Innovative AI Applications in Pharmacology) is a collaborative research group spanning artificial intelligence, deep learning, signal & image processing, and experimental pharmacology. Our projects are led by researchers with complementary expertise in vision transformers, multimodal learning, biomedical signal analysis, and pharmacometrics.

Primary affiliations

• KTÜ ILAFAR — Drug and Pharmaceutical Technology Application & Research Center
• Trabzon University — Department of Artificial Intelligence Engineering
• Collaborative ties with national and international partners in academia, clinics, and industry

Research focus

• Multimodal Pharmacology AI
  Integrating ultrasound & microscopy images, clinical text/metadata, and time-series contractility signals to model drug effects, safety, and efficacy.

• MoA & DDI Modeling
  Learning mechanism-of-action representations and drug–drug interaction (DDI) risks using transformer-based architectures, cross-attention, and correlation-matrix reasoning.

• Biomedical Imaging & EDOF
  Vision-transformer approaches for Extended Depth of Focus (EDOF), shape-from-focus, and task-driven image fusion in microscopic systems.

• Domain Generalization & Robustness
  Methods that generalize across devices, centers, and cohorts, with rigorous ablations and calibration analysis.

Approach & toolset

• Architectures: ViT/PVT/Swin variants, CNN-Transformers, BERT/clinical language models, cross-attention fusion, correlation-matrix reasoning
• Data: biomedical images (ultrasound, microscopy), contractility time-series, curated clinical descriptors
• Practices: reproducible pipelines, cross-validation, stratified evaluation, lightweight deployment

Facilities

We maintain GPU-accelerated training infrastructure and secure data pipelines for clinical collaboration. Our wet-lab partners provide controlled experimental protocols and validated annotations to support model development.

Open collaboration

We actively collaborate with clinicians, pharmacologists, and industry R&D teams.
Interested in joining as a student, visiting scholar, or partner lab? Reach out at hulya@ktu.edu.tr with a short bio and interests.

Selected highlights

• PVT-EDOF for Microscopy — Vision-transformer pipelines for extended depth-of-focus reconstruction and optimal image fusion.
• Contractility Signal Intelligence — 1D/2D deep models on pharmacodynamic time-series and Gram-based transforms for MoA-aware readouts.

Ethics & data governance

We follow institutional and national ethics guidelines, emphasize privacy-preserving methods, and report dataset characteristics, limitations, and potential biases. Where applicable, IRB/ethics approvals and informed-consent procedures are observed.

This page auto-lists selected publications and recent posts if available in your repository. To feature works, mark entries in _bibliography/papers.bib with selected={true}; news items go in _news/, and posts in _posts/.