Automatic generation of primary sequence patterns from sets of related protein sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Protein Science

Background:

• Homologous protein families share conserved sequence elements.
• Identifying these conserved patterns is crucial for understanding protein function and evolution.
• Existing methods may not fully capture the nuances of conserved patterns.

Purpose of the Study:

• To develop a novel computer algorithm for extracting conserved primary sequence patterns from homologous protein families.
• To create a library of these patterns for protein sequence databases.
• To evaluate the diagnostic power of these patterns for sequence family membership.

Main Methods:

• Clustering pairwise similarity scores of related protein sequences to form a binary dendrogram.
• Stepwise reduction of the dendrogram by generating common patterns at each node using an extended dynamic programming algorithm.
• Utilizing a nested hierarchy of amino acid classes to define patterns and a "pay once" gap penalty rule.

Main Results:

• Successfully developed a computer algorithm to generate "covering patterns" for homologous protein families.
• Created a library of these patterns for the National Biomedical Research Foundation/Protein Identification Resource database.
• Demonstrated that covering patterns are more diagnostic for sequence family membership than individual sequences.

Conclusions:

• The developed algorithm effectively extracts conserved sequence patterns.
• Covering patterns offer improved accuracy in identifying homologous protein family members.
• This approach enhances sequence analysis and database querying capabilities.