To deal with repeated data or longitudinal data, linear mixed effects models are commonly used. A classical parameter estimation method is the Expectation–Maximization (EM) algorithm. In this paper, we propose three new Partial Least Square (PLS) based approaches using the EM-algorithm to reduce the high-dimensional data to a lower one for fixed effects in linear mixed models. Unlike the Principal Component Regression approach, the PLS method allows to take into account the link between the outcome and the independent variables. We compare these approaches from a simulation study and a yeast cell-cycle gene expression data set. We demonstrate the performance of two of them and we recommend their use to conduct future analyses for high dimensional data in linear mixed effect models context.