Search Results

You are looking at 1 - 3 of 3 items for

  • Author: Igor Tom x
Clear All Modify Search
Open access

Inese Polaka, Igor Tom and Arkady Borisov

Decision Tree Classifiers in Bioinformatics

This paper presents a literature review of articles related to the use of decision tree classifiers in gene microarray data analysis published in the last ten years. The main focus is on researches solving the cancer classification problem using single decision tree classifiers (algorithms C4.5 and CART) and decision tree forests (e.g. random forests) showing strengths and weaknesses of the proposed methodologies when compared to other popular classification methods. The article also touches the use of decision tree classifiers in gene selection.

Open access

Natalia Novoselova, Igor Tom and Michael Belevtsev

Abstract

This article describes the software and underlined decision support models for the immunophenotype diagnostics of leukosis (leukemia) and lymphomas adjusted for the marker or human leukocyte antigen (CD-antigen) coexpressions. Using the model knowledge base, the decision inference algorithm allows computing the degree of manifestation of the disease subtypes for the input immunophenotype features. Software provides the twostage diagnostics of the leukemia subtypes and the lymphoma diagnostics using the set of the developed rules, possibility to observe the diagnostic results and corresponding reference information. The patient data are organized according to the unified registration card, which provides the possibility to work at the different diagnostic levels: diagnostics of the extended groups of leukosis, diagnostics of the leukemia subtypes, diagnostics of the adult and child lymphomas.

Open access

Natalia Novoselova, Igor Tom, Arkady Borisov and Inese Polaka

Abstract

This article considers the gene ranking algorithm for the microarray data. The rank vector is estimated by classifications of the random data samples. At each iteration, the ranks of genes participating in the successful classification become higher. Unlike other methods of feature selection, the proposed algorithm allows increasing the generality of the classification models by construction of the balanced training samples and taking into account the descriptiveness of the gene combinations by the subset estimation.