MkDom2

MPI_MKDOM2 (the parallelized implementation of MKDOM2) is a high-performance bioinformatics tool for decomposing protein sequence databases into homologous domains and clustering these domains into families, a core prerequisite for protein sequence interpretation and for building large-scale domain resources such as PRODOM. Because domain recognition and clustering procedures typically scale roughly quadratically with database size, MPI_MKDOM2 addresses the resulting computational bottleneck by distributing the MKDOM2 workflow across multiple processors using MPI.

The parallelization is designed to handle key difficulties inherent to the original iterative algorithm: dependencies between iterations that constrain naive task parallelism, highly heterogeneous runtimes across iterations that can cause severe load imbalance, and the risk of communication bottlenecks due to large data volumes. Large-scale evaluations indicate that MPI_MKDOM2 remains robust under uneven task durations while maintaining efficient coordination, enabling PRODOM-scale domain-family construction at sizes that would be impractical with the sequential MKDOM2 by several orders of magnitude.