Predicting Binding Regions within Disordered Proteins

Area of Science:

• Protein biochemistry
• Computational biology
• Structural biology

Background:

• Intrinsically disordered proteins (IDPs) lack fixed structures but perform vital biological functions.
• Disordered regions are implicated in various cellular processes, including molecular binding.
• Predicting function in disordered proteins remains challenging due to their dynamic nature.

Purpose of the Study:

• To evaluate the efficacy of neural network predictors, originally trained on X-ray data, in identifying functional regions within NMR-characterized disordered proteins.
• To explore the potential for different "flavors" or sub-classes of disorder to explain prediction accuracy.

Main Methods:

• Application of neural network disorder predictors to protein sequences previously identified as disordered by Nuclear Magnetic Resonance (NMR).
• Analysis of prediction outputs, specifically false negatives, for correlations with known functional regions.
• Examination of 5 disordered proteins containing 9 functional segments.

Main Results:

• Neural network predictors successfully identified or partially identified 8 out of 9 functional segments across 5 disordered proteins.
• False negative predictions of order in disordered regions often corresponded to functionally important sites.
• Identified functional regions were involved in binding DNA, RNA, and other proteins.

Conclusions:

• Neural network predictors show promise in identifying functional binding regions within intrinsically disordered proteins.
• The accuracy may stem from distinct sequence-based "flavors" of disorder, potentially indicating local tendencies towards order.
• This approach could aid in characterizing binding sites in proteins that are difficult to study structurally.