Computational Structural Biologist | Protein Design | Machine Learning for Biology
London, UK
Email: recepadiyaman2244@gmail.com
Portfolio: https://recep2244.github.io/portfolio/
LinkedIn: https://www.linkedin.com/in/recep-ad%C4%B1yaman/

I am a structural bioinformatician with deep expertise in protein structure prediction, protein-ligand interaction modeling, immune repertoire modeling, and model quality assessment. My work has been benchmarked in blind assessments such as CASP,and CAMEO,and has contributed to tools used in both academic and industry pipelines, including collaborations with InstaDeep and BioNTech.

Multi-chain structure prediction and assembly benchmarking (CASP15-17, CAMEO)

• Combines domain parsing, AF3 recycling, Boltz2, and Chai-1 mechanisms
• Outperformed AlphaFold-Multimer v3 in CAMEO
• MultiFOLD3 under active blind testing in CASP17
• Interface-aware scoring using ICS/iPTM

Ligand binding site and affinity predictor (CASP16)

• Hybrid of template-based and template-free modeling
• Ranked Top 2 for MPscore (pose) and Top 10 for affinity
• Generates ligand conformers, binding poses, and EC/GO function annotations

MD-based protein model refinement (CASP14)

• Ranked Top 5 overall and Best server globally
• Incorporates OpenMM and GROMACS simulations with stereochemical optimization
• Guides refinement using local model accuracy scores

Protein-protein and immune interaction scoring (BioNTech x InstaDeep)

• Over 16,000 structural and energy-based features
• Integrates DockQ, PatchQS, CNN-I, interface lDDT, ICS/iPTM
• Adds FoldX, Rosetta, and AMBER energies for orthogonal signal fusion
• Discriminated strong vs weak TCR-pMHC binders in mutation analyses

• Programming: Python, Bash, PyTorch, scikit-learn, Snakemake
• Structure: AlphaFold, Modeller, Rosetta, FoldX, OpenMM, GROMACS
• Docking and Scoring: PatchDock, DockQ, CNN-I, iPTM/ICS
• Data: Biopython, RDKit, Pandas, NumPy, SciPy
• Visualization: PyMOL, ChimeraX, VMD

• multifold: AF3-based quaternary structure modeling
• funfold: Ligand binding prediction and function annotation
• refold: Protein refinement via MD simulations
• modfolddock2q: Interface quality scoring engine

• House_Price_Prediction: California housing regression baseline with feature engineering and XGBoost.
• titanic-survival-prediction: Titanic survival classification with engineered features and model comparison.
• predict_nvda: NVDA price prediction baseline with market features and time-aware validation.

• Postdoc - University of Reading (McGuffin Group)
• Lead Scientist - InstaDeep (BioNTech collaboration)
• Developer - FunFOLD, ReFOLD, MultiFOLD, ModFOLDdock2Q
• Top-tier performance in CASP14-17, CAMEO, and immune-modeling pipelines

Interested in joint research, tools integration, or ideas at the intersection of machine learning and structural biology? Reach out anytime.