High throughput profile-profile based fold recognition for the entire human proteome

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• Maintaining comprehensive structural annotation databases requires regular updates using advanced profile-profile fold recognition methods.
• The computational demands of these methods necessitate efficient resource management for timely proteome annotation.
• Keeping pace with rapidly updating sequence and structure databases is crucial for high-quality structural modeling.

Purpose of the Study:

• To introduce and benchmark the JYDE (Job Yield Distribution Environment) system, a meta-scheduler for managing intensive computational tasks.
• To demonstrate JYDE's capability in distributing genomic-scale fold recognition workloads across independent Grid domains.
• To accelerate the structural annotation of the Human proteome using the latest mGenTHREADER software and JYDE.

Main Methods:

• Development of JYDE, a meta-scheduler operating above existing cluster schedulers (e.g., SGE, Condor).
• Benchmarking JYDE's performance in distributing intensive profile-profile fold recognition jobs.
• Utilizing the latest mGenTHREADER software for annotating the Human proteome against up-to-date sequence and structure databases.

Main Results:

• The JYDE system effectively scales to manage numerous intensive fold recognition jobs across multiple independent computer clusters.
• JYDE enabled the annotation of 99.9% of Human proteome sequences in under 24 hours.
• This was achieved by leveraging over 500 CPUs distributed across 3 independent Grid domains.

Conclusions:

• The study confirms the feasibility of on-demand, high-quality structural annotation for eukaryotic proteomes.
• JYDE facilitates regular, comprehensive updates of profile-profile based fold recognition models for entire proteomes.
• Grid middleware like JYDE is essential for efficient and timely proteome-wide structural annotation.