PICNIC web server for predicting proteins involved in biomolecular condensates.

MOTIVATION: Biomolecular condensates have been implicated in key cellular processes such as gene regulation, stress response, and signaling, and dysregulation of condensates has been linked to neurodegeneration and other diseases. Computational algorithms that predict protein condensation can aid systematic characterization of biomolecular condensates at the proteome scale. However, many experimental labs may lack the computational background or resources to run sophisticated prediction tools locally. RESULTS: Here, we developed the web server implementation of the PICNIC (Proteins Involved in CoNdensates In Cells) machine learning algorithm. PICNIC uses sequence- and structure-based features derived from AlphaFold2 models to predict if a protein is involved in biomolecular condensates. In case of well-studied proteins with available annotations, the user can further benefit from an extended model, PICNIC-GO, which includes additional features based on Gene Ontology terms. Benchmark tests show that PICNIC algorithms predict condensate forming proteins with ∼80% accuracy. By providing an easy-to-use web server, researchers, without specialized expertise, can rapidly test hypotheses about any protein of interest, including designed and mutated sequences. AVAILABILITY AND IMPLEMENTATION: The PICNIC webserver is available at https://picnic-bio.org/.