In the recent decade, the development of bioinformatics algorithms for feature selection and classification based on the new biomarker platform called Printed Glycan Arrays (PGA) has become one of challenging researched topics. Since many of PGA signals are routinely highly correlated and the data matrices with a high column correlation can adversely affect the results, it is important to take into account this property of PGA chips. The goal of this thesis is to develop a new algorithm inspired by the ideas which followed the development of Correlation Adjusted Wilcoxon ranking (CAW) algorithm, called Combined Correlation Feature selection (CCF). The new algorithm combines correlated features in training set into new uncorrelated features which are subsequently used in feature selection and classification. The thesis presents the implementation of CCF algorithm and its analysis and testing on real data. As shown, the results of the presented experiments suggest a better performance of the CCF algorithm over CAW, and especially over straight-forward univariate feature selection without rejection of correlated features (SW), under certain properties of the correlation structure of data.