ProSOM

ProSOM predicts core promoter regions by applying self-organizing maps to DNA base stacking energy-derived structural profiles to support genome annotation.

Key Features:

• Unsupervised Clustering Approach: Uses self-organizing maps (SOMs) for unsupervised clustering to distinguish structural profiles of promoter sequences from other genomic regions.
• Structural Profile Analysis: Analyzes base stacking energy-derived structural profiles and compares average profiles across transcribed, promoter, and intergenic sequences to identify promoter-specific features.
• Balanced Performance Metrics: Achieves a balanced ratio between predicted promoter sites and false predictions, contributing to robust prediction performance.
• Validation Scheme: Incorporates an objective and biologically grounded validation scheme for core promoter prediction.
• High Precision in Predictions: Validation on ENCODE regions of the human genome showed that 98% of predictions correlate with transcriptionally active regions.

Scientific Applications:

• Genome Annotation: Improves annotation of core promoter locations and transcription initiation sites in genomic sequences.
• Experimental Targeting: Guides experimental design by prioritizing functionally relevant promoter regions for follow-up studies.
• Functional Genomics: Supports gene expression studies and regulatory network analysis by providing high-confidence promoter predictions.

Methodology:

Computational methods include self-organizing maps for unsupervised clustering, calculation of base stacking energy-derived structural profiles, comparison of average structural profiles across transcribed, promoter, and intergenic sequences, and validation using ENCODE human genome regions.