Biomolecular condensates are membraneless organelles that can concentrate hundreds of different proteins to operate essential biological functions. However, accurate identification of their components remains challenging and biased towards proteins with high structural disorder content with focus on self-phase separating (driver) proteins. Here, we present a machine learning algorithm, PICNIC (Proteins Involved in CoNdensates In Cells) to classify proteins involved in biomolecular condensates regardless of their role in condensate formation. PICNIC successfully predicts condensate members by identifying amino acid patterns in the protein sequence and structure in addition to the intrinsic disorder and outperforms previous methods. We performed extensive experimental validation in cellulo and demonstrated that PICNIC accurately predicts 21 out of 24 condensate-forming proteins regardless of their structural disorder content. Even though increasing disorder content was associated with organismal complexity, we found no correlation between predicted condensate proteome content and disorder content across organisms. Overall, we applied a novel machine learning classifier to interrogate condensate components at single protein and whole-proteome levels across the tree of life ( picnic.cd-code.org ).
Intrinsically disordered regions (IDRs) are structurally flexible protein segments with regulatory functions in multiple contexts, such as in the assembly of biomolecular condensates. Since IDRs undergo more rapid evolution than ordered regions, identifying homology of such poorly conserved regions remains challenging for state-of-the-art alignment-based methods that rely on position-specific conservation of residues. Thus, systematic functional annotation and evolutionary analysis of IDRs have been limited, despite comprising ∼21% of proteins. To accurately assess homology between unalignable sequences, we developed an alignment-free sequence comparison algorithm, SHARK (Similarity/Homology Assessment by Relating K-mers). We trained SHARK-dive, a machine learning homology classifier, which achieved superior performance to standard alignment in assessing homology in unalignable sequences, and correctly identified dissimilar IDRs capable of functional rescue in IDR-replacement experiments reported in the literature. SHARK-dive not only predicts functionally similar IDRs, but also identifies cryptic sequence properties and motifs that drive remote homology, thereby facilitating systematic analysis and functional annotation of the unalignable protein universe.