Abstract: The knowledge of the spatial conformation of a protein can help the study of its function, but the number of resolved structures is still limited by the low throughput of the methods used. Structure prediction could bridge the sequence-structure gap, but no reliable and general methods have yet been proposed. An attempt to simplify the problem has been made by trying to predict the contact map of a protein instead of its atoms positions. It has been demonstrated the protein structure can be reconstructed with sufficient precision even if the contact map contains error. Unfortunately, the prediction of contact maps is still very unreliable and it is not clear whether the type of errors made by the predictor can be corrected by the reconstruction method. A low-detail representation of the protein conformation could extract the relevant information to train more efficient predictors. The coarse-grain contact map is defined using contacts between secondary structure segments. The prediction of this type of contacts has been tried, but no results exists about the feasibility of a reliable method that uses only this type of information to reconstruct the protein structure. In this work we concentrate on contacts defined by beta partners. The geometry and connectivity of beta strands imposes strong constraints on the overall structure of the protein, especially for those chains thar are formed mainly by residues in beta conformation. The reconstruction of the structure of this kind of proteins would be enhanced by the knowledge of the secondary structure and the indication of which strands are partners. We propose here an efficient procedure to find a structure that matches the aforementioned characteristics of a given protein in its native conformation.

Abstract: Prediction of protein secondary structure (SS) is a classic problem in computational molecular biology and one of the first successful applications of machine learning to bioinformatics. Most available prediction methods use feedforward neural networks whose input is the multiple alignment profile in a sliding window of residues centered around the target position. By construction, predictions obtained with these methods are local. Long-range dependencies, on the other hand, clearly play an important role in this problem. In it was proposed the use of bidirectional recurrent neural networks (BRNN) for the prediction of SS. The architecture in this case allows us to process the sequence as a whole and to “translate” the input profile at each position into a corresponding output prediction for that position. Theoretically, the output at any position in a BRNN depends on the entire input sequence and thus a BRNN might actually exploit long-range information. Unfortunately, well known problems of vanishing gradients do not allow us to learn these dependencies. In this paper, we are interested in developing an architecture that can effectively exploit long-range dependencies assuming some additional information is available to the learner. We start from a rather simple intuitive argument: if the learner had access to information about which positions pairs are expected to interact, its task would be greatly simplified and it could possibly succeed. In the case of SS prediction, a reasonable source of information about long-range interaction can be obtained from contact maps (CM), a graphical representation of the spatial neighborhood relation among amino acids. Of course in order to obtain a CM the protein structure must be known. In addition, it is well known that backbone atoms’ coordinates can be reconstructed starting from CMs. Thus, in a sense, using CM information in order to predict SS might appear foolish since most of the information about the 3D structure of the protein is already contained in the map. However, the following considerations suggest that this setting is worth investigation: • Algorithms that reconstruct structure from CMs are based on a potential energy function with many local minima whose optimization is not straightforward. Thus it is not clear that a supervised learning algorithm can actually learn to recover SS from CMs. • CMs can be predicted from sequence or can be obtained from structures predicted by ab-initio methods such as Rosetta. Although accuracy of present methods is certainly not sufficient to provide a satisfactory solution to the folding problem, predicted maps may still contain useful information to improve the prediction of lower order properties such as the SS. • Even if CMs are given, the design of a learning algorithm that can fully exploit their information content is not straightforward. For example, Meiler and Baker have shown that SS prediction can be improved by using information about inter-residue distances. Their architecture is a feedforward network fed by average property profiles associated with amino acids that are near in space to the target position. In this way, relative ordering among neighbors in the CM is discarded. The solution proposed in this paper is based on an extended architecture that receives as an additional input a graphical description of the pairwise interactions between sequence positions. We call this architecture interaction enriched BRNN (IEBRNN). Its details are presented in a longer version of this paper.

Abstract: Missing

Abstract: Missing

Abstract: Knowledge about the subcellular localization of proteins can provide important information about their function. A reliable automatic classification method for predicting subcellular localization from sequences is therefore a valuable tool to shed light on protein function and may help towards the solution of genomic scale level problems such as the identification of pharmaceutical targets.

Abstract: Non-free cysteines that are not involved in the formation of disulfide bridges are very often bound to prosthetic groups that include a metal ion and that play an important role in the function of a protein. The discrimination between the presence of a disulfide bridge (DB) or a metal binding site (MBS) in correspondence of a bound cysteine is often a necessary step during the NMR spectral assignment process of metalloproteins and its automation may significantly help towards speeding up the overall process. Several proteins are known where both situations are in principle plausible and it is not always possible to assign a precise function to each cysteine (see e.g. {2,1,5]). We formulate the prediction task as a binary classification problem: given a non-free cysteine and information about flanking residues, predict whether the cysteine can bind to a prosthetic group containing a metal ion (positive class) or it is always bound to another cysteine forming a disulfide bridge (negative class). Firstly, we suggest a nontrivial baseline predictor based on PROSITE pattern hits. Secondly, we introduce a classifier fed by multiple alignment profiles and based on support vector machines (SVM)[3]. We show that the latter classifier is capable of discovering the large majority of the relevant PROSITE patterns, but is also sensitive to signal in the profile sequence that cannot be detected by regular expressions and therefore outperforms the baseline predictor.

Abstract: Missing