Protein_Contact_Network_Explorer-PCNE-

Protein Contact Network Explorer (PCNE)

Python Streamlit NetworkX License: MIT Status

PCNE (Protein Contact Network Explorer) is an interactive web-based tool for constructing, visualising, and analysing Protein Contact Networks derived from PDB structure files. It supports three node representation modes which are Cα, Cβ, and Side-chain Centroid and computes key topological metrics including node degree, clustering coefficient, and betweenness centrality with z-score based classification of residues into Peripheral, Structural Hub, Bottleneck, and Global Critical Hub categories. Community detection is implemented via the Leiden algorithm with per-community DSSP secondary structure composition reporting. Visualisations include an interactive 2D contact network, adjacency and distance heatmaps, and an embedded 3D NGL Viewer with a toggle between betweenness centrality gradient and community membership colouring. Networks can be exported in SIF format for downstream Cytoscape integration.

Live Application: https://lactdr5rfibhg9m5tmamwg.streamlit.app/

Table of Contents

Screenshots

Input Panel Full 2D Contact Network
Input panel with node representation selector Full 2D contact network
Adjacency Matrix Distance Heatmap
Binary adjacency map Pairwise distance heatmap
Hub Scatter Plot Leiden Community Detection
Degree vs betweenness centrality classification Leiden community detection with DSSP composition
NGL Viewer Betweenness NGL Viewer Community
3D structure coloured by betweenness centrality 3D structure coloured by community membership

Features

Methodology

Network Construction

Each residue is represented as a node. Edges are defined by Euclidean distance between residue coordinate representations. A binary adjacency matrix is constructed using a user-defined cutoff rc:

\[d(i,j) = \sqrt{(x_i - x_j)^2 + (y_i - y_j)^2 + (z_i - z_j)^2}\] \[A(i,j) = \begin{cases} 1 & d(i,j) \leq r_c \\ 0 & d(i,j) > r_c \end{cases}\]

PCNE supports three node representation modes. The figure below illustrates how the measured inter-residue distance differs across representations:

Node representation modes

Figure 1: Node representation modes. Cα uses the alpha carbon, Cβ uses the beta carbon, and Side-chain Centroid uses the mean position of all non-hydrogen side-chain heavy atoms. Distance is measured between the respective reference points of neighbouring residues.

For the Side-chain Centroid mode, the centroid position is computed as:

\[\vec{r}_i^{\,SC} = \frac{1}{N_i} \sum_{k=1}^{N_i} \vec{r}_{i,k}\]

where $N_i$ is the number of side-chain heavy atoms and $\vec{r}_{i,k}$ is the coordinate vector of the k-th heavy atom. Glycine, which has no side-chain heavy atoms, falls back to its Cα position.

Topological Metrics

PCNE computes four graph-theoretic metrics per residue: degree (number of contacts), local clustering coefficient (local cohesiveness), Wasserman-Faust normalised closeness centrality (global reachability), and betweenness centrality (information flow bottleneck score).

\[C_{between}(i) = \sum_{s \neq i \neq t} \frac{\sigma_{st}(i)}{\sigma_{st}}\]

Residues are classified into four functional categories using z-score thresholds applied jointly to degree and betweenness distributions:

Category Degree Betweenness
Global Critical Hub High High
Structural Hub High Low
Bottleneck Low High
Peripheral Low Low

Hub scatter plot

Figure 2: Residue classification by degree and betweenness centrality into four functional roles.

Community Detection

Community structure is identified using the Leiden algorithm optimising modularity Q. Partition quality is reported as modularity Q directly within the interface.

\[Q = \frac{1}{2m} \sum_{i,j} \left[ A_{ij} - \frac{k_i k_j}{2m} \right] \delta(c_i, c_j)\]

Each community’s biological relevance is assessed through per-community DSSP secondary structure composition, reporting the fraction of helix (H), strand (E), and coil (C) residues per cluster.

Community detection

Figure 3: Leiden community detection with per-community secondary structure composition bars.

3D Visualisation

The embedded NGL Viewer renders the protein backbone in cartoon representation coloured either by log-normalised betweenness centrality using a coolwarm gradient, or by Leiden community membership using a consistent qualitative colour palette.

\[c_i^{\text{norm}} = \frac{\log(1 + BC_i) - \log(1 + BC_{\min})} {\log(1 + BC_{\max}) - \log(1 + BC_{\min})}\]
NGL betweenness NGL community
Betweenness centrality gradient Leiden community membership

Figure 4: 3D structure viewer with toggle between centrality and community colouring.

Installation

git clone https://github.com/akhuuu2303/Protein_Contact_Network_Explorer-PCNE-
cd Protein_Contact_Network_Explorer-PCNE-
pip install -r requirements.txt
streamlit run app.py

Requirements

Usage

  1. Upload a PDB file or select a demo protein from the provided list
  2. Select node representation (Cα, Cβ, or Side-chain Centroid) and set the contact distance threshold
  3. Explore the network views, community detection, 3D structure viewer, and export results in SIF format for Cytoscape

Citation

If you use PCNE in your research, please cite:

@article{ganapathy2026pcne,
  title   = {Protein Contact Network Explorer: Topological Analysis 
             of Protein Structures},
  author  = {Ganapathy, Akhurath and Krishnan, Sanjana V 
             and Emerson, Arnold},
  journal = {Frontiers in Bioinformatics},
  year    = {2026}
}

Acknowledgements

The authors thank Vellore Institute of Technology (VIT), Vellore, for providing the infrastructure to support this work.

This site is open source. Improve this page.