文摘
High-throughput technologies such as microarray data are a great resource for studying and understanding biological systems at a low cost. However noise present in the data makes it less reliable, and thus many computational methods and algorithms have been developed for removing the noise. We propose a novel noise removal algorithm based on Fourier transform functions. The algorithm optimizes the coefficients of the first and second order Fourier functions and selects the function which maximizes the Spearman correlation to the original data. To demonstrate the performance of this algorithm we compare the prediction accuracy of well known modelling tools, such as network component analysis (NCA), principal component analysis (PCA) and k-means clustering. We compared the performance of these tools on the original noisy data and the data treated with the algorithm. We performed the comparison analysis using three independent real biological data sets (each data set with two replicates). In all cases the proposed algorithm removes the noise in the data and substantially improves the predictions of modelling tools.