Prediction of Liquid-Liquid Phase Separation

Liquid-liquid phase separation (LLPS) is the basis for the formation of membraneless organelles (MLOs) or biomolecular condensates, which play a key role in cellular function. LLPS in biology is thought to be essentially driven by multivalent interactions between molecules, which may occur in proteins between multiple folded structural domains or mediated by intrinsically disordered regions (IDRs). Despite tremendous progress in understanding protein LLPS, the prevalence and distribution of phase-separated proteins remains unclear. Therefore, the development of computational methods to predict phase-separated proteins is important for a deeper understanding of the biological functions of LLPS.

Customized Services

We have different theoretical and computational methods to probe the phase behavior of intrinsically disordered proteins (IDPs) or proteins with intrinsically disordered regions (IDRs). These methods have greatly facilitated customer research on key physical principles and sequence features that are critical for phase separation of IDP or IDR-containing proteins. However, these prediction tools for phase-separated proteins are based on small samples and specific protein features, limiting their application.

Here, our team of experts is dedicated to developing big data-based prediction tools with a wide range of applications. CD BioSciences offers powerful machine learning methods to predict liquid-liquid phase separations. Predictive models can be trained by integrating aspects of protein features, including physical or chemical properties of residues or sequence contexts, as descriptors or vectors.

To create more general tools for phase separation protein prediction, we provide rich databases of phase-separated proteins containing experimentally obtained data. In these databases, the deposited proteins are validated by LLPS experiments. Each database provides basic information about the recorded proteins, as well as their structural and functional annotations. Information on the phase behavior of proteins is also stored in each database in more or less detail.

Advantages of Machine Learning Methods for Predicting LLPS

• Direct and more general prediction of proteins that undergo LLPS.
• Performing better than common phase-separated proteins prediction tools in identifying new phase-separated proteins.
• Based on cross-validation and comparison experiments.
• Can identify both fractional and temporal information.
• Identifying novel scaffolding proteins for stress particles.
• Predicting phase separated protein candidates in the human genome for further study.

These databases can provide valuable information about the LLPS system. They overlap each other to varying degrees. At the same time, each database is designed for a specific purpose and has unique features. If you have any special requirements for our services, please feel free to contact us. We are looking forward to working together with your attractive projects.