Frequently more is known about some of the sequences to be aligned than others. For example, if two or more proteins in the family of interest have had their three-dimensional structures solved by x-ray crystallography, then they may be aligned by inspection of the three dimensional models with much greater confidence than by using sequence information alone. MULTALIGN therefore allows one or more extra sequences to be added to a preexisting alignment.
At a lesser level, certain residues may be expected to align on functional grounds - for example binding studies may suggest that a particular Histidine residue is involved in the catalytic mechanism of two or more of the proteins to be aligned. In this situation it is important to be able to include this information in the alignment. This is particularly important if the overall similarity between the sequences is low since it is likely to lead to a biologically more reasonable alignment.
Certain regions of the sequence(s) may be expected to experience fewer gap events than others. For example the core secondary structural elements (beta strands and alpha helices) generally do not tolerate large insertions or deletions. This feature can be encoded when using MULTALIGN by specifying a variable gap-penalty. (See Barton and Sternberg 1987a for a discussion of the benefits of secondary structure dependent gap-penalties).