Proteome-scale measurement of peptide partitioning into reconstituted condensates reveals that amino acid composition, rather than specific sequence motifs, primarily determines condensate formation of intrinsically disordered regions, with experimentally derived amino acid-specific partition free energies representing fundamental physicochemical interactions that quantitatively predict phase separation behavior across diverse IDRs and correlate with established computational models of disordered protein behavior.

Intrinsically disordered protein regions facilitate cellular organization through phase separation into biomolecular condensates. However, the molecular interactions driving this process remain poorly understood because of experimental limitations. Here, we advance experimental throughput by several orders of magnitude by developing Condensate Partitioning by mRNA-Display (CPmD). The method allows analysis of partitioning of hundred thousand peptides derived from the disordered proteome into reconstituted condensates. Our results demonstrate that the amino acid content, rather than specific sequence, primarily determines partitioning behavior. Importantly, quantification of the partitioning energies of peptides allows us to decipher the ‘molecular grammar’ of the relevant interactions, allowing accurate prediction of the formation of condensates of diverse full-length disordered protein regions. The results reveal how physicochemical properties of disordered regions encode biological functions through formation of biomolecular condensates.