A new method to missing value imputation for immunosignature data

Abstract

The immunosignature technology uses microarray chips of random amino acid sequence peptides to detect diseases based on the change in the profile of circulating antibodies. Diseases are detected using classification algorithms trained on a reduced sample of immunosignature patterns of patients with known diagnoses. The aim of the study was to develop a new method of missing value imputation in immunosignature data, which allows maintaining sufficient accuracy of classification. Materials and Methods. The study was carried out using immunosignature data obtained by utilizing a high-resolution peptide microarray chip with nearly ten thousand peptide cells. The applicability of various missing value imputation methods such as simple imputation, weighted k-nearest neighbors and machine learning techniques (linear regression, random forest, gradient boosting) was evaluated. Results. Missing value imputation method based on gradient boosting has been developed in the framework of the study. Its operating principle implies iterating through all features (attributes) and training on examples (samples) whose values are present in the feature, followed by clarification of missing feature values. This process is repeated until the total training error for all features stops decreasing or until the maximum number of iterations is reached. The root mean squared error is employed as a training error metric. To assess the quality of missing value imputation, classification results based on the data obtained after imputation procedure are used in our investigation. The proposed missing value imputation algorithm based on linear gradient boosting proves to be effective under conditions of a high proportion of missing values as compared to other methods under consideration. The results of the investigation demonstrate the viability of using machine learning techniques for missing value imputation in immunosignature data.