An improved scoring method for protein residue conservation and multiple sequence alignment.

One of the most fundamental operation in biological sequence analysis is multiple sequence alignment (MSA). Optimally aligning multiple sequences is an intractable problem; however, it is a critical tool for biologists to identify the relationships between species and also possibly predict the structure and functionality of biological sequences. The most fundamental step of assembling MSA results is identifying the best location to place the sequence residues. And the accuracy of the sequence assembly depends heavily on the reliability of a scoring function used. With an appropriate scoring function, an MSA program can boost its accuracy of multiple sequence alignment up to 25%. In this study, we present a new, fast, and biologically reliable scoring method, hierarchical expected matching probability (HEP), to use in protein multiple sequence alignment. The new scoring method eliminates the burden of gap cost selection process. And it has consistently proven to be more biologically reliable than all other tested scoring methods through all tests on four different theoretical and experimental benchmarks, Valdar's theoretical conservation benchmark, RT-OSM, BAliBASE3.0, and PREFAB4.0. An implementation of our new scoring method into progressive multiple sequence alignment, resembling the alignment algorithm in PIMA, ClustalW, and T-COFFEE, has shown an accuracy improvement up to 7% on BAliBASE3.0 and up to 5% on PREFAB4.0 benchmarks.