MultiDomainBenchmark: a multi-domain query and subject database suite.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein domain analysis is crucial for understanding protein function and evolution.
• Existing benchmarks for sequence searching tools primarily use single-domain protein queries.
• Most proteins, especially in Metazoa, are multi-domain, exhibiting complex architectures and functions.

Purpose of the Study:

• To introduce a novel benchmark, MultiDomainBenchmark (MDB), for evaluating sequence searching tools.
• To address the limitations of single-domain benchmarks in reflecting real-world protein complexity.
• To assess the performance of commonly used tools like BLAST/PSI-BLAST and HMMER on multi-domain proteins.

Main Methods:

• Developed MultiDomainBenchmark (MDB), a database of 412 multi-domain queries and 227,512 target sequences.
• Included sequences from over 5108 species and 1123 protein families with domain location annotations.
• Introduced a new classification technique for multi-domain proteins to evaluate domain architecture recovery.

Main Results:

• Evaluated the performance of BLAST/PSI-BLAST and HMMER using the MDB.
• Assessed the ability of algorithms to accurately recover domain architectures in multi-domain proteins.
• Demonstrated the necessity of multi-domain benchmarks for comprehensive tool evaluation.

Conclusions:

• MultiDomainBenchmark (MDB) provides a gold standard for evaluating sequence searching tools on complex, multi-domain proteins.
• The MDB facilitates more accurate performance assessments compared to single-domain benchmarks.
• MDB is publicly accessible for researchers to utilize and advance sequence analysis.