Protein sequences predictive methods using

Genome Technology Branch National Human Genome Research Institute National Institutes of Health Bethesda, Maryland 

This chapter focuses on computational techniques that allow for hiological discovery based on the protein sequence itself ov on their comparison to protein families. Unlike nucleotide sequences, which are composed of four bases that are chemically rather similar (yet distinct), the alphabet of 20 amino acids found in proteins allows for much greater diversity of structure and function, primarily because the differences in the chemical makeup of these residues are more pronounced. Each residue can influence the overall physical properties of the protein because these amino acids are basic or acidic, hydrophobic or hydrophilic, and have straight chains, branched chains, or are aromatic. Thus, each residue has certain propensities to form structures of different types in the context of a protein domain. These properties, of course, are the basis for one of the central tenets of biochemistry that sequence specifies conformation (Anflnsen et al., 1961). 

Rather than using an amino acid sequence to search SWISS-PROT, AACompIdent uses the amino acid composition of an imknown protein to identify known proteins of the same composition (Wilkins et al., 1996). As inputs, the program requires the desired amino acid composition, the isoelectric point (pi) and molecular weight of 

For each sequence in the database, the algorithm computes a score based on the difference in compositions between the sequence and the query composition. The results, returned by E-mail, are organized as three ranked lists  

---

Ofiran, Y. and Rost, B., 2005, Predictive methods using protein sequences. In Baxevanis, A.D. and Ouellette, B.F.F. (Eds), Bioinformatics A practical guide to the analysis of genes and proteins, 3rd Edition, Wiley-Interscience, 197-221. 

What can be done by predictive methods if the sequence search fails to reveal any homology with a protein of known tertiary structure Is it possible to model a tertiary structure from the amino acid sequence alone There are no methods available today to do this and obtain a model detailed enough to be of any use, for example, in drug design and protein engineering. This is, however, a very active area of research and quite promising results are being obtained in some cases it is possible to predict correctly the type of protein, a, p, or a/p, and even to derive approximations to the correct fold. 

A potentially general method of identifying a probe is, first, to purify a protein of interest by chromatography (qv) or electrophoresis. Then a partial amino acid sequence of the protein is deterrnined chemically (see Amino acids). The amino acid sequence is used to predict likely short DNA sequences which direct the synthesis of the protein sequence. Because the genetic code uses redundant codons to direct the synthesis of some amino acids, the predicted probe is unlikely to be unique. The least redundant sequence of 25—30 nucleotides is synthesized chemically as a mixture. The mixed probe is used to screen the Hbrary and the identified clones further screened, either with another probe reverse-translated from the known amino acid sequence or by directly sequencing the clones. Whereas not all recombinant clones encode the protein of interest, reiterative screening allows identification of the correct DNA recombinant. 

These predictive methods are very useful in many contexts for example, in the design of novel polypeptides for the identification of possible antigenic epitopes, in the analysis of common motifs in sequences that direct proteins into specific organelles (for instance, mitochondria), and to provide starting models for tertiary structure predictions. 

Smith, 1992). All of these can identify an optimal alignment between the query and either a set of previously studied sequences or a pattern of sequence elements identified as common to a set of previously studied proteins. De novo sequence analysis methods have proved less useful. Although there has been some slow progress in predicting a protein s structure from its sequence, no direct functional predictions methods have been developed. 

Evaluation of protein sequence analysis methods based on the use of PSSMs in benchmarking experiments and in a number of test cases shows that these methods are capable of systematically detecting relationships between proteins that previously have been deemed tractable only at the structure-comparison level. Clearly, however, there is still a lot of room for improvement, as many automated procedures missed subtle connections that subsequendy have been revealed on a case-by-case basis, in part thanks to a careful choice of starting points for the PSSM construction. An exhaustive exploration of the sequence space by recursive iterative searching is likely to yield additional, on many occasions unexpected, links between proteins and, in particular, is expected to increase the rate of structure prediction. 

---