Application of Unscented Particle Filtering for Estimating Parameters and Hidden Variables in Gene Regulatory Network

Recent researches on estimation of parameters of gene regulatory networks by differential equations generally based on Kalman Filtering Model, it makes assumptions that the analyzed system is linear. However, gene regulatory networks are obviously non-linear system, so great deviation error will happen. Here we present a method to estimate the parameters and hidden variables of gene regulatory networks based on Unscented Particle Filter. It makes better fitness than Kalman Filtering Model due to free of the premise that the model is linear. By comparison of the estimation result between Unscented Particle Filter and Unscented Kalman Filter on the hidden variables and parameters of Repressilator, advantage of our method on reduction of estimation error is validated. The amount of particles is simultaneously analyzed. Both deficiency and overabundance of particles will weaken the accuracy of estimation, so selection on the moderate amount of particles is significant.