Intellectual Analisys of Data

Setting Up the Nonlinear Model of Experimental Data From DNA Microarrays

E. A. Riabenko<sup>ab

a</sup> The Faculty of Computational Mathematics and Cybernetics of Lomonosov Moscow State University, Moscow, 119991, Russia
^b SNC Bioclinicum, Moscow, 115088, Russia

Abstract: The nonlinear model for data on DNA expression is presented which uses the Langmuir function to describe the probes fluorescence intensities. The method is based on minimization of the loss function from Alpha-Beta Divergence family. To set up the model parameters the publicly available data of a few thousands microarrays were used. Numerical experiments are conducted to choose optimal values of model’s hyperparameters. It is shown that the model fits fluorescence intensities of the probes on microarrays better than the standard linear model and that the obtained expression estimates are more robust.

Key words: microarrays, summarization, Langmuir model, AB-divergence.

UDC: 519.254

Received 02.10.2012, Published 25.10.2012