Application of multivariate autoregressive modelling for analysis of immunologic networks in man (Q1104877)

We tried to apply Akaike's method of multivariate autoregressive modelling for analyzing immunologic networks in man. When applying this approach on clinically obtained chronologic data, autoregressive coefficients were obtained by means of a least squares method. The approach enabled us to compute Akaike's relative power contribution, which expresses the degree of unidirectional contribution from a certain variable to another. Judging from the degree of contribution, thus estimated, of the \(T4^+\) lymphocyte level to the \(T8^+\) lymphocyte level in a patient with systemic lupus erythematodes (SLE), it seems that slower components in the movements of suppressor T-cells are governed almost exclusively by those of helper T-cells, indicating the malfunction of suppressor T-cells in such a patient. Also, by putting the obtained autoregressive coefficients into a state space representation of the system under study, it became possible for us to simulate an impulse response and/or a step response of the system and thereby to understand the abnormality of networks in pathologic conditions. The present approach may provide a standardized method for analyzing immunologic networks in the body.