Wojciech Lesinski

Integration of multiple types of genetic markers for neuroblastoma may contribute to improved prediction of the overall survival

BackgroundModern experimental techniques deliver data sets containing profiles of tens of thousands of potential molecular and genetic markers that can be used to improve medical diagnostics. Previous studies performed with three different experimental methods for the same set of neuroblastoma patients create opportunity to examine whether augmenting gene expression profiles with information on copy number variation can lead to improved predictions of patients survival. We propose methodology based on comprehensive cross-validation protocol, that includes feature selection within cross-validation loop and classification using machine learning. We also test dependence of results on the feature selection process using four different feature selection methods.ResultsThe models utilising features selected based on information entropy are slightly, but significantly, better than those using features obtained with t-test. The synergy between data on genetic variation and gene expression is possible, but not confirmed. A slight, but statistically significant, increase of the predictive power of machine learning models has been observed for models built on combined data sets. It was found while using both out of bag estimate and in cross-validation performed on a single set of variables. However, the improvement was smaller and non-significant when models were built within full cross-validation procedure that included feature selection within cross-validation loop. Good correlation between performance of the models in the internal and external cross-validation was observed, confirming the robustness of the proposed protocol and results.ConclusionsWe have developed a protocol for building predictive machine learning models. The protocol can provide robust estimates of the model performance on unseen data. It is particularly well-suited for small data sets. We have applied this protocol to develop prognostic models for neuroblastoma, using data on copy number variation and gene expression. We have shown that combining these two sources of information may increase the quality of the models. Nevertheless, the increase is small and larger samples are required to reduce noise and bias arising due to overfitting.ReviewersThis article was reviewed by Lan Hu, Tim Beissbarth and Dimitar Vassilev.

Modeling in Feature and Concept Spaces: Exclusion Relations and Similarities of Features Related with Exclusions

The paper introduces definitions of exclusion relations in spaces of features and concepts. Concepts correspond to phenomena and they are described with their features. The objective of our research is to investigate and describe possible structuring and relations in the feature and concept spaces. In this article, three types of exclusions: weak, strict, and multiple qualitative are defined and discussed. The focus is on similarity of excluding structures. Two distinct methodologies are introduced. First approach relies on direct features comparison. Second methodology uses dedicated similarity relation rooted in the underlying concept space.

Optical Music Recognition: Standard and Cost-Sensitive Learning with Imbalanced Data

The article is focused on a particular aspect of classification, namely the issue of class imbalance. Imbalanced data adversely affects the recognition ability and requires proper classifier’s construction. In this work we present a case of music notation as an example of imbalanced data. Three classification algorithms - random forest, standard SVM and cost-sensitive SVM are described and tested. Feature selection based on random forest feature importance was used. Also, feature dimension reduction using PCA was studied.

Modelling Human Cognitive Processes

The article presents an application of fuzzy sets with triangular norms and balanced fuzzy sets with balanced norms to decision making modelling. We elaborate on a vector-based method for decision problem representation, where each element of a vector corresponds to an argument analysed by a decision maker. Vectors gather information that influence given decision making task. Decision is an outcome of aggregation of information gathered in such vectors. We have capitalized on an inherent ability of balanced norms to aggregate positive and negative premises of different intensity. We have contrasted properties of a bipolar model with a unipolar model based on triangular norms and fuzzy sets. Secondly, we have proposed several aggregation schemes that illustrate different real-life decision making situations. We have shown suitability of the proposed model to represent complex and biased decision making cases.

Imbalanced pattern recognition: Concepts and evaluations

In this paper we propose and investigate a concept of imbalanced pattern recognition problems and evaluation methods of solutions applied to solve such problems. The attention is focused on so called paper-to-computer technologies, but it is not limited to them due to possible direct generalization to other domains. Besides bringing a concept of imbalanced pattern recognition problem, classification quality from the perspective of single classes is considered. Parameters of binary classification and parameters and measures used in signal detection theory are adopted. Quality of classification in terms of one class contra all others is taken into account. Then, classifiers performance in frames of one class at the background of other classes and in frames of impact of other classes on the given on are evaluated. Finally, parameters characterizing global properties of classification are introduced and illustrated.

Decision Trees and Their Families in Imbalanced Pattern Recognition: Recognition with and without Rejection

Decision trees are considered to be among the best classifiers. In this work we use decision trees and its families to the problem of imbalanced data recognition. Considered are aspects of recognition without rejection and with rejection: it is assumed that all recognized elements belong to desired classes in the first case and that some of them are outside of such classes and are not known at classifier’s training stage. The facets of imbalanced data and recognition with rejection affect different real world problems. In this paper we discuss results of experiment of imbalanced data recognition on the case study of music notation symbols. Decision trees and three methods of joining decision trees (simple voting, bagging and random forest) are studied. These methods are used for recognition without and with rejection.