Work mail: foteini.aktypi@sund.ku.dk

Photo of Foteini Aktypi

Research focus

Development and application of deep learning models in biomedical research. I am currently applying machine learning and deep learning models with the focus on the transformer architecture to electronic health record data. My interests revolve around the application of deep learning for the integration of EHRs with omics data to aid biological discoveries.

Current projects:

  • Predicting weight gain for patients receiving the antipsychotic medication olanzapine with machine learning.
  • Generating patient embeddings from sequences of medical events to predict future events and stratify patients with similar medical backgrounds.

Future projects

collaboration with Marc Pielies Avelli - open for collaborations

  • Building omics knowledge graphs and training with deep learning models for biological pathway and cross omics relationship discoveries

Datasets

  • sp-data (Gentofte and RegionH Psychiatry Data)
  • PhenoData 10K Cohort (Weizmann Institute)

Toolkit:

  • EIR
  • DuckDB
  • Pytorch (Python)

Past research:

  • Transcriptomics data analysis (single-cell rna seq, bulk-rna seq, micro-arrays)
  • Whole exome sequencing data analysis
  • Biological network analysis (Cytoscape)
  • Molecular structure analysis (Pymol, Maestro)

Keywords: Electronic Health Record Databases, Machine Learning, Deep Learning, Transformer Neural Networks, Large Language Models.