The efficiency of cross-linked enzyme aggregates (CLEAs) is mainly affected by the strength and binding site of the formed linkages between the enzyme and cross-linker. Therefore, this study investigated the impact of different macromolecular cross-linkers on various functional groups, their binding energy, and intermolecular interaction in generating CLEAs of endolevanase from Bacillus lehensis G1 (rlevblg1), through the combination of computational and experimental analysis. Due to the distanced bonding of dextran from the active site, rlevblg1 cross-linked with dextran (rlevblg1-dex-CLEAs) exhibited the highest binding affinity (− 7.1 kcal/mol) and activity recovery compared to six other cross-linkers. Thus, the role of computational cross-linker screening is confirmed as a crucial step to predict strong attachment and construct efficient CLEAs.