Holley, H.L. and Karplus, M., Protein secondary structure prediction with a neural network, Proc. Natl. Acad. Sci. USA, 86:152--156, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of these approaches use the leave one out criterion, to the best of our knowledge, none of them test performance with the leave half out criterion. Various machine learning techniques have been applied to the protein structure prediction problem. The two main approaches are neural nets (e.g. [47, 67, 59]) and hidden Markov models (e.g. 53, 9] Both of these approaches require adequate data on the target motif, since there is a training session on sequences that are known to contain the target motif. Our approach differs from these methods since it does not require well analyzed data on the ....

L. Holley and M. Karplus. Protein secondary structure prediction with a neural network. Proceedings of the National Academy of Sciences, 86:152--156, 1989.

....secondary structure of proteins can be described as a learning problem as follows: We are given examples of proteins with known primary and secondary structure. Given the primary structure of a target protein, predict its secondary structure. Another formulation of the learning problem is that in [10, 6, 3]: Given a sequence of residues of fixed length (a window ) from a protein chain, classify the middle residue in the window as ff helix, fi sheet, or coil. Several methods have been proposed. Three of the most common are those of Robson, 4, 5] Chou and Fasman, 2] and Lim, 9] Recently, two new ....

....chain, classify the middle residue in the window as ff helix, fi sheet, or coil. Several methods have been proposed. Three of the most common are those of Robson, 4, 5] Chou and Fasman, 2] and Lim, 9] Recently, two new approaches have been presented. The first one uses neural networks, [6, 10] and the other uses instance based learning (IBL) algorithms, 3] All of the above methods only consider the relationship between amino acids and the local secondary structure. This means that they only consider the influence on 1108 the secondary structure of an amino acid by others that are ....

L. H. Holley and M. Karplus. Protein secondary structure prediction with a neural network. Proceedings of the National Academy of Sciences of the United States of America, 86:152--156, 1989.

....of 13 outputs from the first network. The second network still has a hidden layer of 40 units and an output layer of three units. With the cascaded networks, they achieved an accuracy of 64.3 . This means that the learned system is correct in predicting secondary structures 64.3 of the time. Holley and Karplus (1989) independently used a very similar neural network approach. Based on the evidence of high statistical correlation with secondary structure and 8 amino acids on either side of a prediction point (Garnier et al. 1978) the input layer has 17 groups, each with 21 units. The hidden layer has only 2 ....

....60 Concept Learning PLS1 or ID3 53.7 Seshu et al. 60.6 Cost Salzberg 71.0 Exemplar based Learning The classification of the nearest instance in the memory becomes the prediction of the new instance. The classification is then adjusted according to the minimal sequence length restrictions used by Holley and Karplus (1989). These restrictions state that a sheet must span at least two amino acids and a helix must span at least four, as mentioned in Section 4.1.1. The highest accuracy reported was 71.0 with a window size of 19. 4.1.3 Summary of Prediction Accuracy The first three methods in Table 3 are ....

Holley, L. H. and Karplus, M. (1989). Protein secondary structure prediction with a neural network. Proc. Natl. Acad. Sci. USA, 86:152--156.

....percent. The limited accuracy of the predictions is believed to be due to the relatively small number of proteins from which the conformational parameters are calculated. Also, secondary structure might be influenced by the tertiary structure, which is not taken into account at all in this method [HK89] Despite its limitations, the method is very popular and widely used. Many improvements, which we will not discuss here, have been proposed and implemented. However, the original algorithm still seems to be used the most. 3.2.2 Neural Networks A neural network is a numerical model of many ....

....neural network, Qian and Sejnowski carefully selected 106 proteins. A subset of these was used to train the network; the rest were used to test the network. This resulted in a performance of about 64 percent correct predictions. Based on these experiments, various other neural networks were tried [HK89] KCL90] SLX92] but none of them was able to substantially improve the performance. This seems to be due to the same reasons as the Chou Fasman algorithm: the limited size of the training set and the influence of tertiary structure on secondary structure. Research is still going on in order ....

L. Howard Holley and Martin Karplus. Protein Secondary Structure Prediction With a Neural Network. Proceedings National Academic Science USA, 86:152--156, January 1989.

.... 3 Domains We chose for our comparisons three domains that have received considerable attention from the machine learning research community: the word pronunciation task (Sejnowski and Rosenberg, 1986; Shavlik et al. 1989) the prediction of protein secondary structure (Qian and Sejnowski, 1988; Holley and Karplus, 1989), and the prediction of DNA promoter sequences (Towell et al. 1989) Each domain has only symbolic valued features; thus, our MVDM is applicable whereas standard Euclidean distance is not. Sections 3.1 3.3 describe the three databases and the problems they present for learning. 3.1 Protein ....

....then, is: given a sequence of residues from a fixed length window from a protein chain, classify the central residue in the window as ff helix, fi sheet, or coil. The setup is simply: window . TDYGNDVEY z XGQVT E z GTPGKSFNLNFDTG. central residue Qian and Sejnowski (1988) and Holley and Karplus (1989) formulated the problem in exactly the same manner. Both of these studies found the optimal window size to be approximately 17 residues (21 was the largest window tested in either study) In a separate statistical study, Cost (1990) found that a window of size five or six is nearly sufficient for ....

[Article contains additional citation context not shown here]

Holley, L. and Karplus, M. (1989) Protein Secondary Structure Prediction with a Neural Network. Proceedings of the National Academy of Sciences USA, 86, 152-156.

....e.g. Monge et al. 213] Pattern recognition. There have been a number of attempts to predict the secondary structure assignment directly from the sequence of amino acids, using pattern recognition techniques (often neural networks) trained on proteins with known geometric structure; see, e.g. [100, 108, 144, 169, 177, 242, 245, 316]. The many traditional statistical techniques for pattern recognition see, e.g. Fukunaga [102] Young Fu [355] have hardly been tried. However, success was rather limited; see Schulz [282] and Stolorz et al. 309] for critical evaluations of the state of affairs in 1988 and 1991, but also ....

L. Holley and M. Karplus, Protein secondary structure prediction with a neural network, Proc. Natl. Acad. Sci. USA 86 (1989), pp. 152--156.

....to speed up the convergence of the training process. The binary representation is similar to the one used in the NETtalk network (Sejnowski Rosenberg, 1987) which assigned phonemes to alphabetic letters of English text and the one used in networks (Qian Sejnowski, 1988; Bohr et al. 1988; Holley Karplus, 1989; Kneller et al. 1990) which classifies amino acids into categories of their corresponding protein secondary structure: alpha helix, beta sheet and coil. A similar sparse coding scheme was applied in a network predicting distance constraints on the three dimensional structure of protein backbones ....

Holley, L.H. & Karplus, M. (1989). Protein Secondary Structure Prediction With a Neural Network. Proc. Natl. Acad. Sci. 86, 152-156.

....trees. 1.5.3 Secondary structure prediction Chou and Fasman [Chou and Fasman, 1974] and Lim [Lim, 1974] proposed the secondary structure prediction problem almost twenty years ago. Recently, researchers have used artificial intelligence techniques to attack the problem [Qian and Sejnowski, 1988, Holley and Karplus, 1989] Zhang et al. Zhang et al. 1992] describe a system that uses a neural network, called Combiner, to combine the predictions of three experts, a neural network, a memory based reasoning system, and a Bayesian statistical module. Each of the three experts examines a thirteen residue window in ....

....p(N i ) 1 Gamma p(N i ) M ) p(c coil ) 2:4) To extract a number from equation 2. 4 we need to assign numerical values to: M , p(N i ) and p(c coil ) M , the number of instances in the training set, is approximately 10,000 [Zhang et al. 1992] Approximately 54 of the residues are coil [Holley and Karplus, 1989, Kneller et al. 1990] so p(c coil ) 54. Estimating p(N i ) the probability that an instance is in neighborhood i, is more difficult. I assume that instances are evenly distributed, so p(N i ) 1=jN j, where jN j is the number of neighborhoods. Using this approximation, 2.4 reduces to: P ....

[Article contains additional citation context not shown here]

Holley, L. and Karplus, M. (1989). Protein secondary structure prediction with a neural network. Proceedings of the National Academy of Science, 86:152--156.

No context found.

Holley, H.L. and Karplus, M., Protein secondary structure prediction with a neural network, Proc. Natl. Acad. Sci. USA, 86:152--156, 1989.

No context found.

Holley, L.H., Karplus, M. Protein secondary structure prediction with a neural network. Proc. Natl. Acad. Sci. U.S.A. 86:152--156, 1989.

No context found.

, 9893-9908. Holley, L.H. & Karplus, M. (1989). Protein Secondary Structure Prediction With a Neural Network. Proc. Natl. Acad. Sci. 86, 152-156. Jacob, M. & Gallinaro, H. (1989). The 5' splice site: phylogenetic evolution and variable geometry of the association with U1RNA. Nucl. Acids Res. 17,

No context found.

L. H. Holley and M. Karplus, "Protein secondary structure prediction with a neural network," Proc. Natl. Acad. Sci. USA, vol. 86, pp. 152-156, Jan. 1989.

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