Salzberg, S. & Cost, S. (1992). Predicting Protein Secondary Structure with a Nearest-neighbor Algorithm. Journal of Molecular Biology. 227, 371-374.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of secondary structure labels based on the amino acid sequence alone. Present work uses the set of secondary structure labels whose size is 3 (i.e. Helix, Coil, and Sheet) Previous works on predicting secondary structures of proteins have yielded the best percent accuracy ranging from 63 to 71 [8]. These numbers, however, should be taken with caution since performance of a method based on a training set may vary when trained on a different training set. MEMM was recently introduced to model sequential data [6] Some of the examples that this new framework can be applied include tagging ....

....how much each feature function influence in determining the probability of s given x. Roughly speaking, greater the value of a parameter, the more it allows the corresponding feature function to contribute to the final probability value. 5 Experimental Results A data set used in earlier works [8] provided both the training and testing data sets. Briefly, this data set contained amino acid sequences of experimentally solved structures and corresponding sequences of secondary structure labels as determined by the DSSP program [4] The original data set was processed to fit the needs of the ....

[Article contains additional citation context not shown here]

Salzberg, S. & Cost, S. (1992). Predicting Protein Secondary Structure with a Nearest-neighbor Algorithm. Journal of Molecular Biology. 227, 371-374.

....classification. The classifaction of these sequential fragments relies either directly or indirectly on a measure of distance between residue hexamers. 5. 1 Nearest neighbour classification Nearest neighbour classification techniques have been successfully applied to protein structure prediction [19, 18, 25]. These methods are based on the assumption that examples which are closer in the feature space are of the same class. As a measure of distance we use the non Euclidean value difference metric defined by Stanfill and Waltz [23] which relates the distance between two symbolic values to the ....

S. Salzberg and S. Cost. Predicting protein secondary structure with a nearest-neighbor algorithm. J. Mol. Biol., 227:371--374, 1992.

....k DNF formulae. Perceptrons Perceptron learning algorithms with sigmoid threshold functions are the most widely used programs for secondary structure prediction [Qian and Sejnowski, 1988, Holley and Karplus, 1989, Zhang et al. 1992] although other techniques, such as nearest neighbor approaches [Salzberg and Cost, 1992], are now becoming popular. In this section we briefly explore the number of instances needed to train a perceptron with stair step thresholds. Once again we consider a hypothesis to be a set of three perceptrons, one for each of the three types of secondary structure. To train a single perceptron ....

Salzberg, S. and Cost, S. (1992). Predicting protein secondary structure with a nearest-neighbor algorithm. Journal of Molecular Biology, 227:371--374.

....understood for these data sets (and, in fact, some of these datasets are known to be quite easy to classify [Hol93] Therefore, one cannot make any strong conclusions from these results. We primarily wanted to demonstrate that MVDM based systems can be effective for some standard benchmarks. In [SC92, ZMW92] MVDM is used to deliver a relatively high accuracy on protein secondary structure prediction. That is, it matches or exceeds the accuracy of carefully tuned neural networks on a relatively sparse training data which is rather surprising. We now turn to artificial data to better gauge the specific ....

S. Salzberg and S. Cost. Predicting protein secondary structure with a nearest-neighbor algorithm. Journal of Molecular Biology, 227:371--374, 1992.

....in fact, some of these datasets are known to be quite easy to classify [Hol93] Therefore, one cannot make any strong conclusions from these results. We primarily wanted to demonstrate that the MVDM family of algorithms can be effective for some discretized versions of standard benchmarks. In [SC92, XMW92] MVDM is used to deliver a relatively high accuracy on protein secondary structure prediction. That is, it matches or exceeds the accuracy of carefully tuned neural networks on relatively sparse training data, which is rather surprising. We now turn to artificial data to better gauge the specific ....

S. Salzberg and S. Cost. Predicting protein secondary structure with a nearest-neighbor algorithm. Journal of Molecular Biology, 227:371--374, 1992.

....understood for these data sets (and, in fact, some of these datasets are known to be quite easy to classify [Hol93] Therefore, one cannot make any strong conclusions from these results. We primarily wanted to demonstrate that MVDM based systems can be effective for some standard benchmarks. In [SC92, ZMW92] MVDM is used to deliver a relatively high accuracy on protein secondary structure prediction. That is, it matches or exceeds the accuracy of carefully tuned neural networks on a relatively sparse training data which is rather surprising. We now turn to artificial data to better gauge the specific ....

S. Salzberg and S. Cost. Predicting protein secondary structure with a nearest-neighbor algorithm. Journal of Molecular Biology, 227:371--374, 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC