Effect of sequence number on performance
A major improvement of the BRAliBase 2.1 datasets compared to BRAliBase II is the increased range of sequence numbers per set. This allows, for example, to test the influence of sequence number on performance of alignment programs.
It has already been shown that iterative alignment strategies generally perform better than progressive approaches on protein alignments [10]. The same is true for RNA alignments: with increasing number of sequences and decreasing sequence homology iterative programs perform relatively better compared to non-iterative approaches. Figure 2 demonstrates this for PRRN – a representative for an iterative alignment approach – and CLUSTALW as the standard progressive, non-iterative alignment program. The effect is again most notable in the low sequence identity range (APSI < 0.55). In this range, alignment errors occur that can be corrected during the refinement stage of iterative programs. The same can be demonstrated for other iterative vs. non-iterative program combinations like MAFFT or MUSCLE vs. POA or PROALIGN etc. (see supplementary plots on our website [32]).
Figure 2  Performance of Prrn compared to ClustalW in dependence on sequence number per alignment. The plot shows the difference of the scores of PRRN as a representative of an iterative alignment approach and CLUSTALW (standard options) as a representative of a progressive approach.