Piotr Fabian

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.

Developing a new SVM classifier for the extended ES protein structure prediction

This article presents a new SVM classifier for the prediction of the extended early-stage (ES) protein structures. The classifier is based on physicochemical features and position-specific scoring matrix (PSSM). Experiments have shown that prediction results for specific classes are significantly better than those already obtained.

Classification of Protein Interactions Based on Sparse Discriminant Analysis and Energetic Features

Prediction of protein-protein interaction (PPI) types is an important problem in life sciences because of fundamental role of PPIs in many biological processes. In this paper we propose a new classification approach based on the extended classical Fisher linear discriminant analysis (FLDA) to predict obligate and non-obligate protein-protein interactions. To characterize properties of the protein interaction, we proposed to use the binding free energies (total of 282 features). The obtained results are better than in the previous studies.

A method of predicting the secondary protein structure based on dictionaries

Abstract The shape of a protein chain may be analyzed at different levels of details. The ultimate shape description contains three-dimensional coordinates of all atoms in the chain. In many cases, a description of the local shape, namely secondary structure, is enough to determine some properties of proteins. Although obtaining the full three-dimensional (3D) information also defines the secondary structure, the problem of finding this precise 3D shape (tertiary structure) given only the amino acid sequence is very complex. However, the secondary structure may be found even without having the full 3D information. Many methods have been developed for this purpose. Most of them are based on similarities of the analyzed protein chain to other proteins that are already analyzed and have a known secondary structure. The presented paper proposes a method based on dictionaries of known structures for predicting the secondary structure from either the primary structure or the so-called structural code. Accuracies of up to 79% have been achieved.