SuleimanCSB
GitHub

Computational Structural Biology Laboratory

From sequence to atomic mechanism.

Deep-learning structure prediction, vector-algebraic structural analytics and automated docking pipelines — applied to the oncogenic hub kinase AKT1 and the chloroplast thylakoid kinases STN7 / STN8.

Explore research Open 3D explorer
Projects
2
3D models
2
Figures
30
Wiki notes
48

Research portfolio

Projects

Each project is a reproducible pipeline maintained in the private core repository; curated results are published here.

Interactive

3D Structure & Confidence Explorer

Rotate, zoom and inspect the predicted models. Click the pLDDT track to locate a residue in 3D.

Loading 3D engine…
Very high ≥ 90 Confident 70–90 Low 50–70 Very low < 50

Per-residue pLDDT

hover the track

Predicted Aligned Error

Å
0 Å30 Å

Mathematical & biophysical foundations

Methods

Pipelines are built on explicit derivations implemented in Python rather than GUI-only tools.

Local Distance Difference Test (pLDDT)

Superposition-free local agreement over Cα pairs within 15 Å:

$$\text{LDDT}=\frac{1}{4|D|}\sum_{(i,j)\in D}\sum_{t\in\{0.5,1,2,4\}}\mathbb{I}\big(|d^{\text{pred}}_{ij}-d^{\text{true}}_{ij}|\le t\big)$$

Backbone dihedrals (φ, ψ)

From four atoms with bond vectors \(\mathbf v_1,\mathbf v_2,\mathbf v_3\) and normals \(\mathbf n_1=\mathbf v_1\times\mathbf v_2\), \(\mathbf n_2=\mathbf v_2\times\mathbf v_3\):

$$\theta=\operatorname{atan2}\!\Big(\big(\mathbf n_1\times\mathbf n_2\big)\cdot\tfrac{\mathbf v_2}{\lVert\mathbf v_2\rVert},\;\mathbf n_1\cdot\mathbf n_2\Big)$$

Cα contact topology

Euclidean distance matrix with a non-trivial contact definition:

$$D_{ij}=\lVert\mathbf r_i-\mathbf r_j\rVert_2,\qquad D_{ij}\le 8\,\text{Å}\;\wedge\;|i-j|\ge 6$$

Predicted Aligned Error

Expected position error of residue \(j\) when aligned on the frame of residue \(i\):

$$\text{PAE}_{ij}=\mathbb E\big[\lVert\mathbf x^{\text{true}}_j-\mathbf T_i\,\mathbf x^{\text{pred}}_j\rVert\big]$$

Kabsch superposition & RMSD

Optimal rotation from the SVD of the covariance \(H=P^{\top}Q=U\Sigma V^{\top}\):

$$R=V\,\mathrm{diag}(1,1,\det(VU^{\top}))\,U^{\top},\quad \text{RMSD}=\sqrt{\tfrac1N\textstyle\sum_i\lVert R\mathbf p_i-\mathbf q_i\rVert^2}$$

Shannon conservation

Per-column entropy of the multiple sequence alignment (H = 0 ⇒ invariant):

$$H_c=-\sum_{a}p_{a,c}\log_2 p_{a,c}$$

Compounding knowledge

Structural Biology Knowledge Base

An interlinked wiki of entities, concepts, syntheses and source notes compiled alongside the research.

Training

Structural Bioinformatics Curriculum

    About

    Suleiman Hajizadeh

    Structural bioinformatician (Institute of Molecular Biology, IMB) working at the interface of systems biology, deep-learning structure prediction and molecular modelling.

    Research spans the multi-scale path transcriptomics → co-expression network hubs → AlphaFold structure → allosteric targetability, and comparative structural evolution of photosynthetic regulatory kinases.

    Contact Personal GitHub

    Organisation architecture

    • SuleimanComputationalStructuralBiology Private core repository — raw data, MSAs, pipelines, manuscripts, notebooks & the LLM wiki source.
    • SuleimanComputationalStructuralBiology.github.io This public site — curated figures, models and metrics exported by site_export/export_site_data.py.