Liquid–liquid phase separation
Making sense of biomolecular condensates through carefully curated data and more informative benchmarks.
Organisation without a membrane
Cells can concentrate proteins and nucleic acids into assemblies known as biomolecular condensates. Liquid–liquid phase separation is one mechanism that can produce such compartments: interactions among molecules favour a concentrated phase alongside a more dilute one. The behaviour depends on molecular composition and experimental conditions.
For computational studies, the challenge is not simply to decide whether a protein has ever appeared in a condensate. A protein that helps drive phase separation may play a different role from one recruited into an existing assembly. Useful datasets need to preserve those distinctions and the evidence behind them.
Questions we explore
- How should we distinguish proteins that drive phase separation from proteins recruited into condensates?
- Which negative examples make a benchmark informative?
- How well do prediction methods generalise across different datasets?
Our approach
Our work focuses on the data used to study and predict phase separation. We bring together curated protein datasets and examine how the choice of positive and negative examples changes the evaluation of prediction tools.
The aim is to make comparisons easier to interpret and reproduce. Distinguishing drivers, clients and negative examples helps researchers ask more precise questions about what a model has learned, rather than treating every condensate-associated protein as the same kind of example.
Tools, projects & people
Explore our publication catalogue, software and data resources, and the related field of amyloid aggregation.
Publications & research stories
Better data for condensate research
Our curated protein datasets and comparative evaluation of LLPS prediction tools.