Sequence alignments statistical significance

It is equally important that the evaluation of initial assignment of sequences to a superfamily be made from structural rather than functional criteria. This avoids, for example, trivial but important errors from the use of keyword searches to identify members of a superfamily. It also avoids early bias in determining residues/motifs that are conserved across all members of a superfamily. Multiple alignment and motif analysis provide sensitive methods for structure-based confirmation of sequence similarity. Scoring systems that can be applied to such methods provide verification of the statistical significance associated with the comparisons. 

Figure 6.11 Alignment of identities only versus the Blosum 62 matrix. Repeated shuffling and scoring reveal the significance of sequence alignment for human myoglobin versus lupine leghemoglobin with the use of either (A) the simple, identity-based scoring system or (B) the Blosum-62 matrix. The scores for the alignment of the authentic sequences are shown in red. The Blosum matrix provides greater statistical power.

Alexandrov, N. N. and V. V. Solovyev, Statistical significance of ungapped sequence alignments. Pac Symp Biocomput, 1998 p. 463-72. 

Use the default parameters (except the number of alignments and descriptions) such as nr as the database (see Note 3), e-value 10 and the statistical significance threshold to include a sequence for generating the PSSM for the next iteration as 0.005. Change the maximum number of alignments and descriptions, 1000, from the respective pull down menus to retrieve possibly all statistically significant hits. 

Hertz, G. Z. and Stormo, G. D. (1999) Identifying DNA and protein patterns with statistically significant alignments of multiple sequences. Bioinformatics 15, 563-577. 

In the last example, we used RNAalifold essentially to detect structural RNAs within long aligned sequences. The method does not, however, provide a good measure for the statistical significance of predicted consensus structures. For example, it is difficult to decide whether a predicted structure is functionally important or just incidental. 

---