Iterative Search

By means of the performance measure H x) described above, systematic and iterative search for good parameter values can be performed. In this work, one systematic and two iterative search methods were explored. 

The Newton method can also be used to find the extremum (maximum or minimum) of a function/(x), i.e., the value for which the first derivative/ (x) is zero. The iterative search for the extremum is implemented by a formula derived from equation (3.1.29)... 

A generic problem with profile methods that iterate is the possibility of profile wander (also called matrix migration). This occurs when sequences found in early rounds of the iterative search are not found in later rounds of the search. This problem affects both PSI-BLAST and HMMER. This means that one should record all the intermediate steps so that these lost members of the family can be recovered. Profile wander only becomes a problem for large protein families, and therefore the cause of the profile wander may be related to the limits of modeling using profiles. 

Another important aspect of structure prediction using PSSMs is the careful determination of the boundaries of distinct domains whose sequences are used as starting points to construct PSSMs. In iterative database searches, this is critical to avoid inclusion of more than one domain, which results in explosion of an iterative search instead of convergence. Using well-defined domain sequences is particularly important in the analysis of small binding domains that frequently are overlooked if combined with larger, usually enzymatic domains. 

Transitive, iterative searches initiated with the sequences of the anticodon-binding domains of lysyl- and aspartyl-tRNA synthetases provided leads for the identification of biologically interesting, previously unknown OB-fold domains at a statistically significant level (random expectation values <0.01). In particular, OB folds were detected in the eukaryotic replication factor RFA with statistically significant scores this... 

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. 

In Figures 4 and 5, typical three-dimensional plots of the reciprocal of the error function, Eqn.(8), are shown during the iterative search of the optimal alignment of the program for the resolution of the envelope models [1A+2A] and [1B+2B] and for all the combinations of the parameters T[ and Tj in the interval from 120 to 220 (data points or channels). 

The fact that a profile or HMM can pick out new sequences also related to the given family suggests that these should be used to update the profile or HMM used as search pattern. This idea leads to iterative search algorithms where the database is searched repeatedly, each time updating the query pattern with some or all of the newly identified sequences. Psi-Blast [101] is a very successful implementation of this idea. It starts with a single... 

As has been described in Sect. 5.3, the conservation patterns of enzymes are often indicative of the particular family they belong to and can be used for their classification. However, the iterative searches and multiple alignment methods used for their establishment require a certain bioinformatic infrastructure as well as some experience with these issues. If the goal of the analysis is not the detection of novel enzyme families, but rather the classification of a novel sequence into one of the already existing enzyme families, there are a number of protein domain and motif databases that will be useful in this respect[60 61. These databases do not store the sequences themselves but rather work with descriptors of protein families and protein domains. These descriptors can consist of the Profiles or Hidden Markov Models mentioned above, but other types are also being used. With a particular... 

The resulting hypotheses can be used to iteratively search chemical databases Accelrys, Inc. http //www.accelrys.com/products/catalyst/catalystproducts/cathypo.htm... 

From the standpoint of statistics, the transformation Eq. 2.2-19 into Eq. 2.3.b-l and the determination of the parameters from this equation may be criticized. What is minimized by linear regression are the (residuals) between experimental and calculated y-values. The theory requires the error to be normally distributed. This may be true for r, but not necessarily for the group /(Pa - PrPs/I Wa and this may, in principle, affect the values of k, K, K, Ks, — However, when the rate equation is not rearranged, the regression is no longer linear, in general, and the minimization of the sum of squares of residuals becomes iterative. Search procedures are recommended for this (see Marquardt [41]). It is even possible to consider the data at all temperatures simultaneously. The Arrhenius law for the temperature dependence then enters into the equations and increases their nonlinear character. 

The structure query input and query compilation functions were separated from the actual searching function. Query compilation converts the structure input representation as specified by a user into the screen and connection table forms needed by the screen and iterative search processes for the actual searching function. Query negotiation and query compilation were performed on the IBM mainframe computer to take advantage of already existing software. Also, these functions are not computeintensive and so were not an appreciable factor when considering the relative cost of computing resources. 

---