Accurate predictions of disordered protein ensembles with STARLING - PubMed
3 days ago
- #protein-ensembles
- #deep-learning
- #bioinformatics
- STARLING is a framework combining physics-based force fields and generative deep learning to predict disordered protein ensembles.
- It supports environmental conditioning across ionic strengths and allows ensemble refinement under experimental constraints.
- STARLING enables rapid generation of IDR ensembles and ensemble-aware representations from sequence.
- The framework facilitates ensemble-based search for 'biophysical look-alikes' and accelerates sequence design.
- STARLING's latent representations can be used for library-scale designs, reducing computational time significantly.
- It complements bioinformatic protein sequence analysis and aids in hypothesis generation for IDR function.