Summary
I am a Bioinformatics Scientist at the Bioinformatics & Omics Data Science platform of the Max-Delbruck-Center for Molecular Medicine. My roles in the platform include research, collaborations with other labs in the institute, and supporting other researchers in the form of consultations, mentorships, workshops, and user-friendly software development for wet-lab researchers. I am also a member of the arcas.ai team, where my current research is focused on development of deep learning-based multi-modal data integration tools for precision oncology. My work in the last 15 years spanned various categories such as comparative genomics, protein sequence analysis in the context of molecular interactions and disease mechanisms, RNA bioinformatics, and omics data science with a focus on development of reproducible software for the bioinformatics community.
Experience
Bioinformatics Scientist
Max-Delbrück-Center for Molecular Medicine
Berlin, Germany
Postdoctoral Fellow
European Molecular Biology Laboratory
Heidelberg, Germany
Education
PhD in Bioinformatics
European Molecular Biology Laboratory
Heidelberg, Germany
MSc in Bioinformatics
Simon Fraser University, BC Cancer Agency
Vancouver, Canada
BSc in Biological Sciences
Sabanci University
Istanbul, Turkey
Skills
Programming Languages
Active Skills: Python, R, Bash/Shell
Familiar/Past Skills: Perl, C++, SQL
Deep Learning Development
Active Skills: PyTorch, PyTorch Lightning, PyTorch Geometric
Familiar/Past Skills: TensorFlow, Keras
Machine Learning Methods
Random Forests, SVMs, GLMnet
Version Control
Active Skills: Git
Familiar/Past Skills: Subversion
Workflows
Active Skills: Snakemake
Familiar/Past Skills: CWL
Package Management
Active Skills: Conda, Guix
Familiar/Past Skills: Docker
Data Science Toolkits
CRAN/Bioconductor, Python libraries (scikit-learn, NumPy, pandas, Matplotlib)
Keywords
Active: Omic data science for precision oncology, protein language models, causal network analysis, single-cell data analysis
Familiar/Past: Comparative genomics, short linear motifs
Profiles
News Articles
- New textbook for computational genomics (mdc-berlin.de)
- Searching the sewers for viruses (bionity.com)
- AI identifies cancer cells (bionity.com)
- Parallel Genome Editing in Microscopic Worms Maps Regulatory Genomic Elements to Physiology (genengnews.com)
- Stray proteins cause genetic disorders (sciencedaily.com)
- A genetic chaperone for healthy aging? (bionity.com)
Selected Publications
Identifying tumor cells at the single-cell level using machine learning
Ikarus is a machine learning pipeline designed to identify tumor cells from normal cells in single-cell sequencing data. Tested on multiple datasets, it demonstrates high sensitivity and specificity in distinguishing cell types in various experimental contexts.
Parallel genetics of regulatory sequences using scalable genome editing in vivo
The crispr-DART software is used to analyze indel mutations in targeted DNA sequencing, providing insights into the impact of mutations on gene expression and fitness. By applying the software in conjunction with inducible Cas9 and multiplexed guide RNAs, researchers can study regulatory sequences directly in animals, helping to understand their role in health and disease.
Single-cell analyses of aging, inflammation and senescence
This study compiles and analyzes aging-related single-cell gene expression datasets, highlighting unique insights that are difficult to obtain from bulk data. The findings reveal increased cellular senescence markers, inflammatory processes, and transcriptional heterogeneity with age, suggesting single-cell experiments could provide critical information for interventions targeting aging, inflammation, senescence, and disease.
PiGx: reproducible genomics analysis pipelines with GNU Guix
This study introduces a principled approach for building reproducible analysis pipelines and managing their dependencies with GNU Guix, presenting PiGx as a case study. PiGx is a set of highly reproducible pipelines for analyzing various types of sequencing data, generating publication-ready reports and figures, designed for users with command-line experience, potentially benefiting laboratory workers and bioinformaticians alike.
Mutations in Disordered Regions Can Cause Disease by Creating Dileucine Motifs
The study investigates the impact of mutations in intrinsically disordered regions (IDRs) of proteins on protein-protein interactions, finding that mutations in three transmembrane proteins lead to increased clathrin binding and create dileucine motifs. The research reveals that several mutations in disordered regions cause "dileucineopathies," with gained dileucine motifs being significantly overrepresented in structurally disordered cytosolic domains of transmembrane proteins.
RCAS: an RNA centric annotation system for transcriptome-wide regions of interest
RCAS, an R package, is designed to simplify the creation of gene-centric annotations and analysis for genomic regions of interest obtained from various RNA-based omics technologies. With a modular design, flexible usage, and convenient integration options, RCAS can reproduce published findings, generate novel knowledge, and provide contextual knowledge necessary for understanding the functional aspects of biological events involving RNAs.