Irena Roterman-Konieczna

Prediction of Protein-Protein Binding Interfaces

When it comes to regulating protein activity, complexation mechanisms are just as important as ligand binding. Most proteins never exist in isolation – instead they serve as building blocks for more complex systems. Some proteins form multimers to ensure maintain spatial alignment (required e.g. for phase separation in the dual lipid layer and formation of hydrophilic compartments in ion channels (Unwin 2005; Jasti et al.. 2007)); others may require temporary binding of cofactors (e.g. regulation of transcription factors (Huxford et al. 1998)), or are part of complicated protein machinery (e.g. proton-driven rotors in ATP synthases (Boyer 1997; Oster and Wang 1999, 2003)).

Comparative Analysis of Techniques Oriented on the Recognition of Ligand Binding Area in Proteins

This chapter presents an analysis of the various models implemented by software packages which enable computerized identification of ligand binding sites.

Can the Structure of the Hydrophobic Core Determine the Complexation Site

Stabilization of the tertiary protein structure is most often attributed to hydrophobic interactions, although this type of interaction is not specifically reflected in protein force fields. Initial attempts to extend the analysis of traditional nonbinding interactions with factors representing hydrophobic interactions (Levitt 1976) were not particularly successful, even though the influence of the aqueous environment on molecular dynamics cannot be underestimated in respect to experimental observations.

An improved protein fold recognition with support vector machines

Predicting the three-dimensional structure (fold) of a protein is a key problem in molecular biology. It is also interesting issue for statistical methods recognition. In this paper a multi-class support vector machine (SVM) classifier is used on a real world data set. The SVM is a binary classifier, but protein fold recognition is a multi-class problem. So several new approaches to deal with this issue are presented including a modification of the well-known one-versus-one strategy. However, in this strategy the number of different binary classifiers that must be trained is quickly increasing with the number of classes. The methods proposed in this paper show how this problem can be overcome.

The early-stage intermediate

Abstract: The multistep polypeptide chain folding model presented in this chapter involves several intermediates, the first of which is called the early-stage (ES) intermediate. This intermediate is assumed to be defined solely on the basis of the backbone conformation and does not take side chains into account. The geometric principles that guide the backbone alignment process and its quantitative influence on the structural arrangement of the folded chain can be expressed by means of a contingency table, linking known structural motifs to specific polypeptide sequences. The basic unit of this algorithm is the tetrapeptide, and the corresponding ES conformational subspace is assumed to consist of seven types of motifs. This limited subspace represents a subset of the full conformational space (i.e., the Ramachandran plot). The volumetric structure of the ES intermediate corresponds to the output of the early folding stage and, simultaneously, provides input for further stages of the folding process.

Structural information involved in the interpretation of the stepwise protein folding process

Abstract: Calculating the quantity of information present in each step of the protein folding process suggests that the multistep approach requires less information than the one-step model. Quantitative analysis reveals that the amino acids present in the polypeptide chain do not carry enough information to accurately predict the values of the angles Φ and Ψ in folded proteins. This conclusion results from comparing the amount of information carried by amino acids with the quantity of information necessary to determine Φ and Ψ, taking the complete Ramachandran map as the conformational space. It is shown that the two-step model (comprising two stages, the ES and LS) requires less information, owing to the fact that the final predictions of the angles Φ and Ψ can be based on a preexisting ES structure. Analysis based on information theory points to particular zones of the Ramachandran map that appear to play an important role in the context of protein structure prediction.

An Efficient Multi-class Support Vector Machine Classifier for Protein Fold Recognition

Predicting the three-dimensional (3D) structure of a protein is a key problem in molecular biology. It is also interesting issue for statistical methods recognition. In this paper a multi-class Support Vector Machine (SVM) classifier is used on a real world data set. The SVM is a binary classifier and how to effectively extend a binary to the multi-class classifier case is still an on-going research problem. The new efficient approach is proposed in this paper. The obtained results are promising and reveal areas for possible further work.

Forensic voice comparison by means of artificial neural networks

Abstract This article examines the effectiveness of artificial neural networks (ANNs) as forensic voice comparison techniques. This study specifically considers feed-forward multilayer perceptron (MLP) and radial basic function (RBF) network models. Formant frequencies of Polish vowel e (stressed or unstressed) in selected contexts were used as predictors. This has already been confirmed in an earlier investigation that determined that dynamic formant frequencies of vowels are powerful elements in distinguishing the voice. It has been concluded that neural networks might assist in distinguishing speakers from the others with very good accuracy, reaching 100%. MLP models should be given preference. The results of the investigation have shown the influence of vowel e triads on the effectiveness of correct classification rates. In addition, the authors have determined that the accuracy of classification is greater when based on a single context than for similar input data aggregated over several different contexts.

Involvement of medical experts in legal proceedings: an e-learning approach

Abstract E-learning programs based on the “Virtual Patient” paradigm familiarize students with the process of combining information derived from different branches of medical science. In addition, medical practice often requires paralegal knowledge – for example, when determining the degree of disability or taking part in medical malpractice proceedings. This paper serves as an introduction to inclusion of modern IT tools in teaching curricula. Such tools are available to almost every student of medical sciences and frequently employ the “Virtual Patient” concept mentioned above. For the purposes of our study, we have prepared a selection of training materials using the CASUS software. The specific features of our study include involvement in legal proceedings based on a retrospective approach, i.e., reconstruction of past events.

Machine Learning Methods for the Protein Fold Recognition Problem

The protein fold recognition problem is crucial in bioinformatics. It is usually solved using sequence comparison methods but when proteins similar in structure share little in the way of sequence homology they fail and machine learning methods are used to predict the structure of the protein. The imbalance of the data sets, the number of outliers and the high number of classes make the task very complex. We try to explain the methodology for building classifiers for protein fold recognition and to cover all the major results in this field.