MSAprobs

MSAprobs performs protein multiple sequence alignment using hidden Markov models to produce high-accuracy MSAs for phylogenetic analysis, structural prediction, and functional annotation of proteins.

Key Features:

• Hidden Markov models: Uses hidden Markov models (HMMs) to compute protein multiple sequence alignments.
• High alignment accuracy: Produces high alignment accuracy, which is important for downstream phylogenetic, structural, and functional analyses, with increased runtimes on large datasets.
• Parallelized implementation (MSAProbs-MPI): Provides a distributed-memory parallel version that leverages MPI on multicore CPU clusters to reduce execution times.
• Cluster performance: Demonstrated runtime reductions of over one order of magnitude on a 32-node cluster with two Intel Haswell processors per node.
• Comparison to GPU methods: Outperforms the GPU-accelerated QuickProbs running on a Tesla K20 in speed when using eight nodes.
• Scalability: Can handle large datasets that exceed the time and memory constraints of both serial MSAProbs and QuickProbs.
• Implementation: Implemented in C++ with MPI and compatible with Linux systems.

Scientific Applications:

• Phylogenetic analysis: Produces MSAs suitable for phylogenetic tree inference and evolutionary studies.
• Structural prediction: Provides alignments to support protein secondary and tertiary structure prediction and comparative modeling.
• Functional annotation: Generates alignments useful for transferring functional annotations and identifying conserved residues in proteins.

Methodology:

Performs alignment computation using hidden Markov models and implements distributed-memory MPI-based parallelization (MSAProbs-MPI) in C++ for multicore CPU clusters.