Parallelized protein secondary structure prediction

Functional characterization of a protein sequence is one of the most frequent problems in biology. Secondary structure prediction is a useful first step in understanding how the amino acid sequence of a protein determines the native state. There are many previous famous prediction methods available now, although most of them allow users to submit their queries to dedicated servers, bulk protein sequence prediction service such as the whole database sequence prediction is still not widely available or requires tedious waiting tune. The author proposed a parallelized protein secondary structure prediction method namely mpiPSSP, using message passing interface (MPI) on a multi-node compute cluster. A preprocess mpiSEQ processes and converts the fasta-like database to the format preferred by mpiPSSP, and a post process mpiRES collects and reports the result as well. mpiPSSP can self-determine the appropriate number of processors according to the size of the querying database or also run using the user-desired number of processors. As the tasks are made independent of each other, a highly scalable solution is achieved.