QUARTERplus

QUARTERplus: Deep Learning-Based Intrinsic Disorder Prediction with Residue-Level Quality Assessment

QUARTERplus refines intrinsic disorder predictions in protein sequences using a deep learning meta-model that improves predictive accuracy and provides residue-level quality assessment scores.

Key Features:

• Deep Learning Meta-Model: Applies a meta-model to enhance intrinsic disorder predictions beyond traditional predictors.
• Residue-Level QA Scores: Generates quality assessment scores for each amino acid to quantify confidence in disorder predictions.
• Error Correction Mechanism: Uses QA scores to identify and correct inaccurate disorder predictions, increasing overall fidelity.

Scientific Applications:

• Intrinsically Disordered Protein Analysis: Supports proteomics research on intrinsically disordered proteins (IDPs) by identifying reliable disordered regions for experimental design and hypothesis generation.

Methodology:

QUARTERplus was empirically evaluated using a large, independent dataset with low sequence similarity. Performance benchmarking against twelve leading disorder predictors achieved an Area Under the Curve (AUC) of 0.93. QA scores show strong correlation with actual predictive quality, validating residue-level reliability assessment.