Figure 3.

The disordered spacers in fungal Dcp2 and vertebrates FET proteins adopt conformations that promote phase-separation. (A) The FCR can distinguish the conformations of spacer segments in multivalent proteins. Proteins with FCR >0.5 preferentially adopt extended conformations like idealized self-avoiding random coils. Proteins with FCR <0.3 can form compact globules. Sequences with intermediate values of FCR form conformations similar to Flory random coils, where the net attractive and repulsive forces between residues and solvent molecules are in balance. The FCR for (B) fungal Dcp2 sequences, and (C) vertebrate HNRNPA1a, a member of the FET family. (D) Schematic for machine-learning random-forest classification to classify spacer types from their amino acid sequence. In brief, we used the sequences of naturally occurring disordered sequences that connect different domains, calculated the average of 500 amino acid properties for each sequence, and used this dataset to classify these sequences into the two categories of self-avoiding random coils, and Flory-random coils and compact globules. (E) The fraction of spacers that adopt compact conformations (Flory random coils, and compact globules), and those that adopt extended conformations (self-avoiding random coils) in fungal Dcp2 and vertebrate FET proteins.