Intrinsically disordered proteins (IDPs) can promote liquid-liquid phase separation and form biomolecular condensates in cellular organization. Studies showcased that the low-complexity regions of IDPs containing repetitive patterns with limited amino acid diversity impart their phase behavior. To explore the design and engineering space of IDPs for biomolecular condensate formation, we used a data-driven approach to identify the minimal consensus repeat sequences of IDPs and design IDP-like polypeptides to reconstruct their phase behavior. Using a one-hot encoding strategy, we aligned the sequences based on amino acids’ chemical properties and calculated the similarity scores of protein sequences from the DisProt database. We selected four top-candidate sequences over 67 promising ones that show highly repeatable low-complexity regions, extracted the sequence pattern, and designed the corresponding analogs with minimal consensus repeats. Both native and engineered sequences are prepared via expression in Escherichia coli (E. coli) and characterized for solution properties (e.g., zeta potential and secondary structures) and phase behavior. We found that the charge-neutral N-terminal domain of Galectin-3 and its engineered analogs selectively formed coacervates with anionic carbohydrates over other biomolecules. In addition, we also investigate a charged amino acid-rich sequence, a proline-rich sequence, and a tyrosine-containing sequence to understand how different sequences affect the selectivity of coacervate formation. We anticipate this work to provide a novel method to design tunable biomolecular condensates for medicine and bioseparation technologies.
