Sirius PSB: a generic system for analysis of biological sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Computational tools are vital for modern biological research, aiding in tasks like protein homology searches and genome annotation.
• Existing methods, such as BLAST, rely on sequence similarity for identifying related proteins.

Purpose of the Study:

• To introduce Sirius Prediction Systems Builder (PSB), a new computational tool for biological sequence analysis, classification, and searching.
• To present novel approaches in sequence analysis, including feature space exploration with genetic algorithms and similarity searching based on features rather than sequence homology.

Main Methods:

• Sirius PSB offers four main operations: building and deploying classifiers, searching for similar proteins, and performing preliminary/post-prediction analysis.
• It utilizes a genetic algorithm for feature identification within the feature space.
• Protein similarity searches are performed based on feature similarity, distinct from traditional sequence similarity methods.

Main Results:

• Two prediction models were successfully built using Sirius PSB: one for Arabidopsis polyadenylation sites and another for protein subcellular localization.
• These models demonstrated competitive performance against state-of-the-art methods on public datasets.
• Sirius PSB significantly reduced the time and effort required for model development.

Conclusions:

• Sirius PSB is an effective and user-friendly computational tool that streamlines sequence analysis and model building.
• Its novel feature-based similarity search can identify functionally related proteins even when sequence homology methods like BLAST fail.
• The tool offers a valuable advancement for biological research, enhancing the discovery of protein relationships and functional sites.