Research
My research lies at the intersection of computational structural biology, bioinformatics, and AI/ML. I develop multiscale models and data-driven methods to understand how proteins and peptides interact with membranes, solvents, and ligands, and to design new bioactive peptides, biologics, and food proteins with improved functionality.
I work closely with experimental collaborators in food and grain science to connect atomistic simulations, quantum chemistry, and protein language models to measurable properties such as gluten viscoelasticity, allergenicity, and antimicrobial activity.
Core research areas
Research themes
AI-enhanced peptide prediction & design
Protein language models, graph neural networks, and deep learning for peptide bioactivity.
- Developed pLM4CPPs, a protein language model–based CNN predictor for cell-penetrating peptides that outperforms existing CPP predictors and offers interpretable sequence features.
- Building pLM4G-CPPs, a graph learning framework that integrates pLM embeddings and graph neural networks for improved classification and analysis of CPPs and antimicrobial peptides.
- Extending unified deep learning architectures (e.g., UniDL4BioPep) for broader bioactive peptide families and agriculture/food applications.
Membrane-active peptides & membrane biophysics
Mechanisms of selectivity, pore formation, and lipid heterogeneity.
- Elucidated membrane selectivity mechanisms of Snakin-Z across prokaryotic and eukaryotic membrane models using all-atom and coarse-grained MD, contact analysis, and free-energy profiling.
- Investigating concentration-dependent melittin pore formation in different lipid compositions to understand transitions from surface-bound states to transmembrane pores.
- Modeled heterogeneous bilayers to study how lipid composition and nanodomain formation regulate peptide binding, insertion, and transport.
Food protein structure, processing & allergenicity
Linking molecular dynamics to functional properties in food systems.
- Performed microsecond-scale simulations of α-gliadin in ethanol–water mixtures to reveal how solvent composition reshapes structure, flexibility, and solvation, with implications for gluten viscoelasticity and dough behavior.
- Studied structural and allergenicity changes in ovalbumin under high hydrostatic pressure and ultrasound, correlating MD and experimental digestion data.
- Contributed to work on frozen dough, plant proteins, and sorghum/millet proteins, integrating MD insights with rheology, RVA, and baking performance.
Non-covalent interactions & quantum chemistry
Fundamental understanding of cation–π, aromatic–aromatic, and ion–water interactions.
- Developed theoretical descriptors and criteria to characterize non-covalent bonds in diverse chemical and biological environments, clarifying the balance between electrostatics, dispersion, and induction.
- Performed first-principles studies on ion–water microsolvation and competing pathways such as water splitting vs. ion solvation for transition metal ions.
- Provided quantum-chemical insight into cation–π interactions and hydrophobic host–guest binding, informing protein design and supramolecular assembly.
Computational drug discovery & bioinformatics platforms
Web-based toolkits and cheminformatics pipelines for structure-based design.
- Co-developed the Molecular Property Diagnostic Suite (MPDS), including fragment library modules and disease-specific portals for tuberculosis and COVID-19, enabling systematic chemical space analysis and hit prioritization.
- Contributed to the development of A2ID 2.0, the Aromatic–Aromatic Interaction Database, and the Cation–Aromatic Database (CAD) for protein design and structural analysis.
- Led in silico studies on herbicide discovery, host–guest complexes, and drug repurposing using docking, MD, and QSAR/ML approaches.
Selected ongoing projects
pLM4G-CPPs: Graph-based learning for CPPs
Extends pLM4CPPs by representing peptide sequences as graphs that combine residue-level embeddings from protein language models with structural/contact information. The goal is to improve robustness on diverse CPP datasets, quantify uncertainty, and provide interpretable sub-sequence motifs linked to uptake efficiency and toxicity.
Techniques: ESM/ProtT5 embeddings, GAT/GCN models, stratified cross-validation, external benchmarking against KELM-CPPpred and MLCPP 2.0.
Melittin pore formation across membranes
Uses large-scale all-atom MD and enhanced sampling to understand how melittin transitions from surface-bound to transmembrane states in eukaryotic vs prokaryotic membranes, and how peptide concentration, lipid composition, and pressure modulate pore stability and leakage.
Techniques: CHARMM force fields, umbrella sampling, free-energy profiles, contact maps, hydrogen-bond and water-wire analyses.
Antifreeze proteins & ice–water interfaces
Simulates antifreeze proteins (AFPs) at TIP4P/ICE ice–water interfaces to dissect adsorption mechanisms, control of ice growth, and thermal hysteresis. Focuses on basal vs prism plane binding, hydrophobic/hydrophilic patterning, and relevance for frozen food stability.
Techniques: TIP4P/ICE water, CHARMM/OPLS force fields, interface-specific analysis, order parameters and clustering of ice-binding modes.
Food protein functionality from atoms to applications
Bridges molecular simulations of gluten proteins, ovalbumin, plant proteins and process conditions (pressure, ethanol, ultrasound) with experimental readouts from rheology, RVA, and baking performance. The aim is to design next-generation plant-based and wheat-based products with targeted texture, stability, and reduced allergenicity.
Techniques: long-timescale MD, coarse-grained simulations, PCA/FEL analysis, contact networks, and integration with experimental grain quality metrics.
Methods & tools
Molecular modeling & simulation
- All-atom and coarse-grained MD of proteins, peptides, membranes, and ice–water interfaces.
- QM and QM/MM for non-covalent interactions, reaction mechanisms, and ion–water microsolvation.
- Enhanced sampling: umbrella sampling, free-energy landscapes, PCA/FEL, clustering, contact network analysis.
- Structure preparation with AlphaFold/ESMFold, Modeller, and visualization in PyMOL/Chimera.
AI/ML, bioinformatics & software
- Deep learning with CNNs and GNNs for peptide classification, activity prediction, and interaction networks.
- Protein language models (ESM-2, ProtT5, Bepler, SeqVec) for sequence and structure-aware embeddings.
- Cheminformatics workflows for virtual screening, chemical space exploration, and fragment libraries.
- Development of reproducible pipelines, web servers, and databases (MPDS, A2ID, CAD) on Linux/HPC systems.
Collaboration
I enjoy collaborating with experimentalists, computational scientists, and data scientists on problems at the interface of proteins, peptides, and food or health applications. I am particularly interested in:
- Co-designing bioactive peptides (CPPs, AMPs, AFPs) guided by simulations and AI models.
- Integrating MD and experimental measurements for food proteins and plant-based formulations.
- Building open-source tools and datasets for peptide design, non-covalent interactions, and drug discovery.
If you are interested in collaborating or hosting a seminar, please feel free to contact me via email or LinkedIn (links on the home page).