Patterns, profiles and multiple sequence alignment

We have not covered BLAST or FASTA searching in this tutorial because they are not currently part of EMBOSS; these searches are offered at many web sites worldwide. However, database searches are an important part of the bioinformatician's arsenal. When we screen a new sequence against a database of known sequences, we are trying to answer the following questions:

• Is there any protein of known structure that has sufficient similarity to the sequence of the unknown protein to suggest a familial relationship?
• If not, which sequence of any known proteins is most similar to the sequence of the unknown protein?

If we can identify a relationship to a protein of known structure, it is possible to infer that the new protein shares a common structure with its relative and to assign its general fold. However, what if the homologue has no known structure? If its function has been identified then we might expect our unknown protein to have a similar or related function. However, exceptions do exist. A classic example is lysozyme, which shares around 50% sequence identity and 70% sequence similarity with $\alpha$-lactalbumin. The two proteins also share similar folds, but their functions are entirely different: the two key catalytic residues of lysozyme are not conserved in $\alpha$-lactalbumin, and the acidic calcium binding motif important to the function of $\alpha$-lactalbumin is not present in most lysoszymes. It is essential that you confirm any computer based predictions with benchwork.

What can you do if sequence similarity does not identify a relative? In this chapter we will show you a few more applications within EMBOSS that can help you predict the function of your sequence.