Lyophilized fungal mycelia are a promising class of biocatalysts, offering high stability in organic solvents and at high temperatures as well as reduced environmental footprint compared to free enzymes. However, the screening of fungal strains for catalytic activity in ester bond formation remains a labor-intensive and costly process. In this study, two predictive models were developed to estimate the catalytic efficiency of fungal mycelia for the esterification of perillyl alcohol with butyric acid. The models were based on attenuated total reflectance infrared (ATR-IR) spectral data and experimentally determined esterification yields of 123 fungal strains isolated from the local environment. To reduce the high dimensionality of the ATR-IR data set (1762 predictors), elastic net regularization and principal component analysis (PCA) were applied. A generalized linear model (GLM) using an elastic net achieved a coefficient of determination (R2) of 0.86, while the PCA-based model yielded an R2 of 0.64. These findings demonstrate a strong correlation between the ATR-IR spectral features and the biocatalytic performance of the mycelia, indicating that their catalytic activity can be reliably predicted using the spectral data alone. This approach enables the rapid and inexpensive screening of fungal biocatalysts, significantly reducing the need for time-consuming enzymatic reactions and chromatographic analysis. The catalytic efficiency, expressed as the synthesis yield of perillyl butyrate, can thus be estimated within seconds using the model FN(0,s2)(⟨x,θ̂⟩) ∈ [0,1] where ⟨x,θ̂⟩ is the inner product between the spectral data vector and the estimated model parameters.