Computational studyBioinformatics · Docking

Protein Mutation Modeling and G6PD Screening

Two linked computational workstreams: multimodal mutation analysis over sequence and structure, followed by a focused study of how the R166H variant changes G6PD structure and small-molecule binding profiles.

Project record

This page reports the maturity and evidence documented in Ethosoft's working archive. It does not imply peer review, regulatory clearance, or clinical validation.

Methods and evidence

Computational studyBioinformatics · Docking

Protein Mutation Modeling and G6PD Screening

Mutation models: NCBI cDNA and protein data, AlphaFold structures, a GNN, sequence-based XGBoost/ensemble models, and a combined multimodal model. The brief reports F1 scores of 72%, 84%, and 94% respectively.
G6PD case study: Wild-type and R166H G6PD sequences modeled with AlphaFold and compared at the structural level.
Screen: More than 8,000 DrugCentral compounds were filtered with RDKit; 3,799 molecules were docked against both forms with GNINA.
Analysis: Binding energies, mutant-versus-wild-type ΔΔG, CNN pose scores, distributions, and top-ligand comparisons.

Evidence boundary: the F1 values are project-reported and the docking results are computational hypotheses. Neither establishes a therapy or proves restoration of protein function without independent and laboratory validation.

Expanded project brief

Sequence, structure, and virtual screening in one workflow.

The protein project combines two related analyses. First, human cDNA and protein sequences sourced from NCBI are represented through sequence models, AlphaFold-derived structures, and graph neural networks; XGBoost and ensemble baselines are compared with a multimodal model that joins sequence and structural evidence. The project notes report F1 values of 72% for the conventional machine-learning path, 84% for the GNN path, and 94% for the combined model.

The second analysis focuses on the R166H G6PD mutation. Wild-type and mutant structures are compared, compounds are filtered with RDKit, and GNINA is used to compare docking poses and binding-energy patterns. This is useful as a hypothesis-generation workflow: ranking molecules computationally can prioritize experiments, but docking scores do not establish binding, efficacy, or restoration of enzyme function.

Reproducibility priorities: report sequence and structure versions, train/test separation by protein family or variant lineage, docking protocol, protonation assumptions, and laboratory follow-up.

Verbatim project material

Original source text

The following text is reproduced exactly from the supplied project files. It is preserved for provenance; it may be in Turkish, use Markdown notation, or contain draft statements.

protein/protein.txt

Download original file

**Protein İlaç Etkileşimi Araştırması | TÜBİTAK 2204-A**

Bu projede, NIH bünyesindeki NCBI veri tabanından alınan insanlara ait cDNA ve protein dizileri AlphaFold ile üç boyutlu yapılara dönüştürülmüş; bu yapılarla eğitilen GNN modeli, protein dizilerinden mutasyon varlığını tespit eden XGBoost ve ensemble tabanlı makine öğrenmesi modeliyle karşılaştırılmıştır. Ardından iki yaklaşım birleştirilerek multimodal bir model geliştirilmiş ve F1 skorları makine öğrenmesi için %72, GNN için %84, multimodal model için ise %94 olarak ölçülmüştür. Ayrıca mutant proteinlerin DrugBank’teki ilaçlarla etkileşimleri GNINA kullanılarak analiz edilerek bazı ilaçların protein yapı ve işlevlerini normal hâline yaklaştırma potansiyeli araştırılmıştır.