Schematic representation of a cell functional enviromics algorithm. The aim is to maximise the covariance between measured environmental factors (X) and rate of change of environmental factors (R) under the constraint of known genes translated into a plausible set of elementary flux modes (E). As in conventional partial least squares, the input matrix (X) is decomposed into a loadings matrix of latent variables (W) and a scores matrix (T). The response data (R), however, are decomposed into genome dependent factors (the structure of elementary flux modes, E) and envirome dependent factors (weights of elementary flux modes, Λ). Then, only the envirome dependent factors (Λ) are linearly regressed against envirome data (X). Finally, such regression coefficients are organized into a functional enviromics map (FEM) with columns representing EMs, rows environmental factors and each element representing the strength of up- or down-regulation of each core cellular function by each envirome factor. Mathematical details can be found as Methods.