Workload optimization of proteomics pattern matching using embedded accelerator

Digitalization has brought a tremendous momentum to the healthcare research. Recognition of patterns in one protein, which are similar to a functional site of another, is crucial for identifying possible functions of newly discovered proteins, as well as analysis of known proteins for previously undetermined activity. PROSITE is a comprehensive database which describes protein domains, families, functional sites and patterns to identify them. In this paper the workload is the task of locating patterns from the PROSITE database over proteins. We optimize the the workload by using IBM's new Power Edge of Network processor (PowerEN) regular expression (RegX) hardware accelerator, which was built for deep-packet inspection at multiple 10Gbps ports. Our preliminary results demonstrate a speedup of 240 relative to software pattern matching. Moreover, indications show that speedup an order of magnitude higher is achievable.