3D clustering of gene expression data from systemic autoinflammatory diseases using self-organizing maps (Clust3D)

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Systemic autoinflammatory diseases (SAIDs) lack specific biomarkers for diagnosis.
• Gene expression time-series clustering is crucial for understanding SAID dynamics.
• Machine learning aids in discovering novel SAID biomarkers.

Purpose of the Study:

• To propose a novel clustering methodology for associating three-dimensional data.
• To efficiently capture temporal dynamics in gene expression patterns for SAIDs.
• To improve the accuracy of pathway-specific gene extraction from clusters.

Main Methods:

• Developed a novel clustering methodology using competitive learning and self-organizing neural networks.
• Applied the algorithm to high-dimensional, time-dependent feature space.
• Evaluated clustering using standard indices and differential expression analysis.
• Compared performance against a heuristic time-series clustering method.

Main Results:

• The proposed methodology significantly improved clustering indices (e.g., threefold increase in Calinski-Harabasz, twofold in Davies-Bouldin).
• Achieved enhanced classification specificity scores.
• Demonstrated superior performance in extracting pathway-specific genes compared to heuristic methods.

Conclusions:

• A novel clustering methodology was successfully developed and applied to SAID gene expression data.
• The method efficiently produces well-separated clusters, outperforming existing heuristic approaches.
• This advancement aids in identifying crucial biomarkers for systemic autoinflammatory diseases.