Roopal SampatBiochemistry 118QProfessor Doug Brutlag3/8/99Probabilistic Approaches to Predicting the Secondary Structure of ProteinsToday’s increasingly unaffordable medical treatment forces genomicresearch to have far-reaching consequences. Most members of the public do notrealize that the genetic sequence does not only encode information abouthereditary make-up, but that it also contains the necessary blueprints for thestructural formation of essential proteins. As the central dogma of molecularbiology declares, DNA is transcribed into RNA, which is then translated intoamino acids that make up proteins. Malfunctions in these proteins result inphenotypes that may be classified as diseases. As health care functions today,doctors assess symptoms, resulting in a diagnosis for a disease. Physicians mustmake educated guesses based upon the symptoms and run a series of tests, theprocess of which may sometimes prove impractical or extremely expensive.Bioinformatics has emerged as providing a new perspective for the treatment ofgenetically inherited diseases. The central paradigm of bioinformatics states thatgenetic information can be used to predict molecular structure of proteins, andthe function of these proteins can then be determined, providing a cause forsymptoms of a disease. If the structure and function of every protein encoded byDNA were known, the underlying causes of symptoms could be easilypinpointed. Elucidating these structures, however, is a process that couldoccupy scientists for hundreds of years. As a result, much research has been andcontinues to be done regarding the prediction of secondary structures of proteinsbased upon determined amino acid sequences.X-ray crystallography has been the traditional method for determining thestructure of a protein. Protein samples are crystallized, and a fine beam of x-raysis targeted at them. The x-ray diffraction detected is then used to generate amodel of the electron density of the protein. Several disadvantages, however,exist to using x-ray crystallography. First of all, the crystallization of proteins isusually a difficult and time consuming process that requires a great deal of skill.Secondly, x-ray diffraction provides a static model of protein structure, withatoms and molecules mapped in fixed-space. Although this representation isuseful, proteins do not usually acquire a fixed structure and instead arecontinuously bending and shifting, characteristics that may be crucial to thefunction of the protein. Thirdly, the time needed to crystallize and x-ray, muchless identify, every single protein that is encoded by the genetic sequence couldspan centuries of work. As a result, scientists would prefer to be able toaccurately predict structure rather than actually determining it.The prediction of protein structure from the amino acid sequence is awork-in-progress. Scientists are cataloguing and using the known structures ofthousands of proteins to help them through this process. The Protein Data Bank,or PDB, is maintained through Brookhaven National Laboratory. As of March 3,1999, the PDB holds 9419 coordinate entries, of which 8751 are proteins, 656 arenucleic acids, and 12 are carbohydrates (Protein Data Bank). These structures areclassified into groups, the most general of which being the Class (α, β, α/β, andα+β); major structural similarities place proteins in the same Fold category; somedegree of sequence similarity implies a probable common ancestry, and putsproteins in the same Superfamily; and greater than 25 percent sequencesimilarity demonstrates a clear evolutionary ancestry, which places proteins inthe same Family (Brutlag lecture, 2/1). The classification of proteins into suchgroups aids in understanding and attempting to predict protein structures byallowing easy observation of and comparisons between patterns in amino acidsequences.The Asilomar conferences of 1994 and 1996 discussed four approaches tosecondary structure prediction. The first is homology modeling. Two proteinsare generally agreed to have the same structure if their sequences are 25-30percent homologous (Brutlag lecture, 2/1). This approach utilizes knowledge ofa closely related protein to predict the structure of a protein in question. If thesequences and/or structures of no closely related proteins are known, however,ab initio prediction appeals as a second approach. Ab initio methods attempt topredict secondary structure through knowledge of only the amino acid sequenceof the protein in question (Altman lecture). One ab initio method that has beenworked on is determining the lowest energy configuration possible, determinedthrough a hidden Markov models and computer modeling, using the givensequence of amino acids. Such an approach, however, has not proven successfulbeyond predicting the secondary structure of small proteins because naturallyoccurring proteins often do not exist in their minimum energy configuration forreasons that may or may not be known (Brutlag lecture, 2/4). For proteins thathave some (< 25 percent) sequence homology with known structures, a thirdapproach to structure prediction is taken. Fold recognition utilizes knowledge ofexisting structures to hypothesize whether or not new sequences could acquiresuch structures (Altman lecture). Predicting how two proteins fit together isdone by the fourth approach, protein docking. The geometry of the physicalassociation between two proteins is predicted by studying the surface-to-surfaceinteractions to determine the best way in which they would fit together.Homology and ab initio are the two current methods that will be concentrated onas efforts towards protein structure prediction.Most efforts at predicting secondary structure concentrate on predictingthe state of an amino acid in the center of a local window of residues (Schmidler).Because the twenty amino acids do not occur in equal distribution in proteins,the beginnings of structure prediction attempted to utilize the frequency of anamino acid’s occurrence in different conformations. For example, proteinsusually have low levels of methionine and tryptophan and higher levels ofleucine and serine (Stryer). In particular, however, the amino acids do not havethe same proportions in particular regions of a protein forms a secondarystructure as they do in the protein overall. The side chains on the amino acidscan either promote or hinder secondary structure formation. Proline disrupts
View Full Document
Unlocking...