Propose an algorithm for Multiple Sequence Alignment (MSA) of protein sequences that improves upon the progressive alignment method (like ClustalW) by incorporating consistency-based scoring. Discuss the computational complexity of your approach and how it handles gaps and mutations in conserved regions differently from standard dynamic programming approaches.