APOD

APOD predicts disordered flexible linkers (DFLs) in protein sequences to identify intrinsically disordered regions that connect protein domains and mediate disorder-based allosteric regulation.

Key Features:

• Dual-Level Input Utilization: Incorporates both local (window-based) and protein-level sequence inputs to capture disorder propensity, sequence composition, conservation, and structural properties, extending the approach used by DFLpred.
• Support Vector Machine Model: Implements a well-parametrized support vector machine (SVM) that integrates the broader set of local and protein-level features for DFL prediction.
• Performance Metrics: On an independent low-sequence-similarity dataset, achieves area under the curve (AUC) 0.82 (28% improvement over DFLpred) and Matthews correlation coefficient (MCC) 0.42 (180% increase over DFLpred).

Scientific Applications:

• Understanding Allosteric Regulation: Identifies DFLs that may mediate disorder-based allosteric regulation and influence protein activity and interactions.
• Protein Engineering: Supports selection or modification of linker regions for designing synthetic proteins with desired flexibility and domain connectivity.
• Drug Target Identification: Aids identification of potential therapeutic targets within intrinsically disordered proteins where structured binding sites are limited.

Methodology:

Integrates local (window-based) and protein-level sequence features—including disorder propensity, sequence composition, conservation, and structural properties—and applies a parametrized support vector machine, with evaluation performed on an independent low-sequence-similarity dataset.