About

Hi, I’m Igor

I work at the intersection of structural bioinformatics, deep learning, and molecule generation for drug design. I like building models and tools that are actually usable for large-scale virtual screening.

Currently

  • PhD student in computer science at Constructor University (Bremen).
  • Researcher at a startup working on computational methods for drug discovery.
  • Working on binding-site prediction and generative models for small molecules / ligands.

Research interests

  • Deep learning on 3D molecular data (grids, point clouds, graphs).
  • Binding site identification and ion/ligand specificity.
  • Generative models for molecule design (VAEs, diffusion, hierarchical models).
  • Practical pipelines: from PDB datasets to screening-ready models.

Selected work

  • Spatiotemporal identification of druggable binding sites – deep-learning framework to detect transient pockets from MD simulations
    (Communications Biology, 2020, 10.1038/s42003-020-01350-0).
  • Protein–peptide binding site detection – 3D CNN models for protein–peptide interfaces
    (JCIM, 2021, 10.1021/acs.jcim.1c00475).
  • Binding site detection in nucleic acid macromolecules – extension of structure-based deep learning to RNA/DNA binding sites
    (NAR Genomics and Bioinformatics, 2021, 10.1093/nargab/lqab111).
  • Review: computational methods for binding site prediction on macromolecules – overview of current approaches and open challenges
    (Quarterly Reviews of Biophysics, 2025, 10.1017/S003358352500006X).

In the meantime, you can see some of the technical background in my Research Blog.

About this site

This website has two main parts:

  • Research Blog - technical posts about structural bioinformatics, deep learning, datasets, and tooling.
  • Notes - more informal things: music, books, generative art, and various experiments.

The goal is to document things I find interesting or useful enough to write down.

Elsewhere