Muscle v5 is an extensive re-write of the MUSCLE code based on on new algorithms.
Compared to previous versions, Muscle v5 is much more accurate, is often faster, and scales to much larger datasets. At the time of writing (late 2021), Muscle v5 has the highest scores on multiple alignment benchmarks including Balibase, Bralibase, Prefab and Balifam. It can align tens of thousands of sequences with high accuracy on a low-cost commodity computer (say, an 8-core Intel CPU with 32 Gb RAM). On large datasets, Muscle v5 is 20-30% more accurate than MAFFT and Clustal-Omega.
Alignment ensembles
Muscle v5 can generate ensembles of high-accuracy alternative alignments ("replicates"). All replicates
have equal average accuracy on benchmark test, including the MSA made with default parameters. So
default parameters should not be considered recommend parameters, as is often
assumed when biologists use bioinformatics software
Differences between replicates are necessarily due to errors. The typical number of differences between replicates represents a lower bound on the number of errors you should expect to exist in any one of the alignments.
By comparing results of downstream analysis (trees, structure prediction...) on different replicates, you can assess the effects of alignment errors on your study.
