Research Portfolio
Prof. Dr. Petr Popov’s core expertise lies in developing physics-based machine-learning methods for molecular science and computational biology. Over the past fifteen years, he has built numerical approaches that accelerate scientific discovery across structural biology, drug discovery, protein engineering, and computational chemistry. His work spans the complete computational discovery pipeline: geometric and graph neural networks for molecular representation learning, deep learning for protein structure and binding-site prediction, AI-driven virtual screening, large language models for biological sequence analysis, and optimization methods for molecular design. Alongside methodological advances, he has translated these approaches into practical scientific software and patented technologies that are used to identify druggable targets, predict protein stability, design receptor mutations, and support structure-based drug discovery. Petr leads interdisciplinary collaborations spanning machine learning, structural biology, chemistry, and medicine while producing influential publications in leading journals.
What connects these activities – and what Petr brings to Constructor Labs – is the view that future breakthroughs in biology will depend on AI systems capable of both generating new scientific knowledge and communicating that knowledge effectively to human researchers. His work focuses on building AI that reasons across molecular structures, biological mechanisms, and experimental evidence, while producing tools that remain interpretable, reproducible, and directly useful to scientists. This combination of autonomous scientific reasoning, deployable software systems, and educational impact positions him to contribute across Constructor Labs’ research agenda, from AI-driven discovery in life sciences to the development of next-generation AI systems that augment both scientific research and human learning.
AI Agents for Life Science Research
Develops autonomous AI-assisted frameworks that combine LLM-based agents, modular MCP-enabled tools, and virtual experiment engines to streamline wet-lab protein engineering and bioinformatics workflows. The goal is to reduce repetitive manual analysis, enable continuous human–AI collaboration, and accelerate research cycles in areas such as protein thermostabilization, enzyme engineering, and antibody design.
- Crystal Structure of the Human Cannabinoid Receptor CB2
- 5-HT2C Receptor Structures Reveal the Structural Basis of GPCR Polypharmacology
- Prediction of homoprotein and heteroprotein complexes by protein docking and template‐based modeling: A CASP‐CAPRI experiment
- Native phasing of x-ray free-electron laser data for a G protein–coupled receptor
- Community‐wide evaluation of methods for predicting the effect of mutations on protein–protein interactions
- Crystal structure of the Frizzled 4 receptor in a ligand-free state
- graphDelta: MPNN Scoring Function for the Affinity Prediction of Protein–Ligand Complexes
- Structure-based mechanism of cysteinyl leukotriene receptor inhibition by antiasthmatic drugs
- Computational design of thermostabilizing point mutations for G protein-coupled receptors
- Spatiotemporal identification of druggable binding sites using deep learning
- Structural basis of ligand selectivity and disease mutations in cysteinyl leukotriene receptors
- Crystal structure of misoprostol bound to the labor inducer prostaglandin E 2 receptor
- PEPSI-Dock: a detailed data-driven protein–protein interaction potential accelerated by polar Fourier correlation
- Rapid determination of RMSDs corresponding to macromolecular rigid body motions
- Protein–Peptide Binding Site Detection Using 3D Convolutional Neural Networks
- Knowledge of Native Protein–Protein Interfaces Is Sufficient To Construct Predictive Models for the Selection of Binding Candidates
- Molecular mechanism of biased signaling at the kappa opioid receptor
- Prediction of disease-associated mutations in the transmembrane regions of proteins with known 3D structure
- DockTrina: Docking triangular protein trimers
- Mutations in transmembrane proteins: diseases, evolutionary insights, prediction and comparison with globular proteins
- Yaroslav V. Solovev, Nikita N. Kostin, Yuri A. Prokopenko, Patrick Masson, Ivan V. Smirnov, Hongkai Zhang, Wei Zheng, Igor A. Yaroshevich, Alexey V. Stepanov, Petr A. Popov, and Alexander G. Gabibov
Chemical neighborhood exploration for substrate discovery in biocatalysis
In: Proceedings of the National Academy of Sciences, Vol. 123 | No. 24, 123 (24) e2535430123
Date of publication: 8 June, 2026
DOI: https://www.pnas.org/doi/10.1073/pnas.2535430123
Publications on Scopus
- CU Cilia – an application for image analysis by machine learning – reveals significance of cysteine cathepsin K activity for primary cilia of human thyroid epithelial cells
- Computational methods for binding site prediction on macromolecules
- OrgNet: Orientation-gnostic protein stability assessment using convolutional neural networks
- Approaching Optimal pH Enzyme Prediction with Large Language Models
- graphLambda: Fusion Graph Neural Networks for Binding Affinity Prediction
- Unraveling viral drug targets: a deep learning-based approach for the identification of potential binding sites
Aleksandr Shumilov
PhD student / Junior Researcher
Works in structural bioinformatics and chemoinformatics, focusing on the generation and assessment of conformational ensembles and their application in computer-aided drug design. The same AI-driven approach to discovery, combining simulation with automated model building, carries over to materials science.
Anastasia Sarycheva
PhD Student
Research focuses on structural bioinformatics, computational drug discovery, and computational modeling. She is currently working on methods for protein stability and molecular property prediction, and on knowledge-based expert systems for analyzing experimental biochemical data.
Igor Kozlovskii
Researcher
Works at the intersection of structural bioinformatics and deep learning, with a focus on computational analysis of molecular interactions and binding sites. His current research includes machine-learning methods for protein–ligand and RNA–ligand interactions, binding-site prediction, and representation learning for biomolecular structures.
Mark Zaretskii
PhD Student
Research interests lie at the intersection of AI, chemoinformatics, and materials science, with current projects focusing on the development of AI tools and pipelines that can accelerate the identification and prioritization of promising molecules and materials for drug design and neuromorphic electronics, respectively.