FLIPPER: Predicting and Characterizing Linear Interacting Peptides in the Protein Data Bank

Area of Science:

• Structural biology
• Bioinformatics
• Computational biology

Background:

• Many protein regions are disordered in isolation and fold upon binding, playing key roles in signaling and regulation.
• These regions, known as intrinsically disordered proteins (IDPs) or linear motifs, are challenging to detect automatically at the proteome level.
• Current public databases have limited examples of these functionally important regions.

Purpose of the Study:

• To develop an automatic method, FLIPPER, for detecting structurally linear regions or peptides that interact with another chain in a protein complex.
• To improve the identification and characterization of intrinsically disordered protein regions (IDPs) that gain structure upon binding.

Main Methods:

• FLIPPER utilizes a random forest classification approach, taking protein structure as input.
• The method considers various structural features: intra- and inter-chain contacts, secondary structure, solvent accessibility (bound/unbound), linearity, and chain length.
• Models were trained and validated on independent datasets like PixelDB-25 and DIBS-25.

Main Results:

• FLIPPER demonstrates high accuracy, achieving 99% precision and sensitivity on PixelDB-25, and 87-88% on DIBS-25.
• Processing the entire Protein Data Bank revealed diverse classes of LIPs with distinct binding modes.
• A significant portion of identified LIPs were not detected by existing disorder predictors.

Conclusions:

• FLIPPER provides an accurate and automated method for identifying protein regions that fold upon binding (LIPs).
• The identified LIPs offer insights into diverse binding mechanisms and highlight limitations of current disorder prediction tools.
• FLIPPER predictions are integrated into MobiDB 4.0, enhancing the database's utility for studying protein interactions and regulation.